Thread: Re: [mail] Re: Windows Build System
> -----Original Message----- > From: Vince Vielhaber [mailto:vev@michvhf.com] > Sent: 30 January 2003 09:17 > To: Ron Mayer > Cc: pgsql-hackers@postgresql.org > Subject: Re: [mail] Re: [HACKERS] Windows Build System > > > On Wed, 29 Jan 2003, Ron Mayer wrote: > > > > > Cool irony in the automated .sig on the mailinglist software... > > > > On Wed, 29 Jan 2003, Vince Vielhaber wrote: > > > ... > > > hammering the betas is a far cry from an "industrial-strength > > > solution". ... TIP 4: Don't 'kill -9' the postmaster > > > > Sounds like you're basically saying is > > > > _do_ 'kill -9' the postmaster... > > > > and make sure it recovers gracefully when testing for an > "industrial- > > strength solution". > > Not what I said at all. It's not far off, but it's quite amusing none the less. What I read from your postings it that you are demanding more rigourous testing for a new major feature *prior* to it being comitted to CVS in a dev cycle than I think we ever gave any previous new feature even in the beta test phase. I don't object to testing, and have been thinking about coding something to address Tom's concerns, but let's demand heavy testing for the right reasons, not just to try to justify not doing a Win32 port. I would also point out that we already list the Cygwin port of PostgreSQL as supported. Who ever gave that the kind of testing people are demanding now? I think the worst case scenario will be that our Win32 port is far better than the existing 'supported' solution. Regards, Dave.
On Thu, 30 Jan 2003, Dave Page wrote: > > On Wed, 29 Jan 2003, Ron Mayer wrote: > > > > > > > > Cool irony in the automated .sig on the mailinglist software... > > > > > > On Wed, 29 Jan 2003, Vince Vielhaber wrote: > > > > ... > > > > hammering the betas is a far cry from an "industrial-strength > > > > solution". ... TIP 4: Don't 'kill -9' the postmaster > > > > > > Sounds like you're basically saying is > > > > > > _do_ 'kill -9' the postmaster... > > > > > > and make sure it recovers gracefully when testing for an > > "industrial- > > > strength solution". > > > > Not what I said at all. > > It's not far off, but it's quite amusing none the less. I agree with Tom on yanking the plug while it's operating. Do you know the difference between kill -9 and yanking the plug? > What I read from your postings it that you are demanding more rigourous > testing for a new major feature *prior* to it being comitted to CVS in a > dev cycle than I think we ever gave any previous new feature even in the > beta test phase. I don't object to testing, and have been thinking about > coding something to address Tom's concerns, but let's demand heavy > testing for the right reasons, not just to try to justify not doing a > Win32 port. Nice try. I've demanded nothing, quit twisting my words to fit your argument. If you're going to test and call it conclusive, do some conclusive testing or call it something else. But I suspect that since you don't know the difference between yanking the plug and kill -9 this conversation is a waste of time. Vince. -- Fast, inexpensive internet service 56k and beyond! http://www.pop4.net/ http://www.meanstreamradio.com http://www.unknown-artists.com Internet radio: It's not file sharing, it's just radio.
On Thu, 2003-01-30 at 13:24, Vince Vielhaber wrote: > On Thu, 30 Jan 2003, Dave Page wrote: > > > > On Wed, 29 Jan 2003, Ron Mayer wrote: > > > > > > > > > > > Cool irony in the automated .sig on the mailinglist software... > > > > > > > > On Wed, 29 Jan 2003, Vince Vielhaber wrote: > > > > > ... > > > > > hammering the betas is a far cry from an "industrial-strength > > > > > solution". ... TIP 4: Don't 'kill -9' the postmaster > > > > > > > > Sounds like you're basically saying is > > > > > > > > _do_ 'kill -9' the postmaster... > > > > > > > > and make sure it recovers gracefully when testing for an > > > "industrial- > > > > strength solution". > > > > > > Not what I said at all. > > > > It's not far off, but it's quite amusing none the less. > > I agree with Tom on yanking the plug while it's operating. Do you > know the difference between kill -9 and yanking the plug? Kill -9 seems to me _less_ severe than yanking the plug but much easier to automate, so that could be the first thing to test. You have no hope of passing the pull-the-plug test if you can't survive even kill -9. Perhaps we could have a special "reliability-regression" test that does "kill -9 postmaster", repeatedly, at random intervals, and checks for consistency ? Maybe we will find even some options for some OS'es to "force-unmount" disks. I guess that setting IDE disk's to read-only with hdparm could possibly achieve something like that on Linux. > > What I read from your postings it that you are demanding more rigourous > > testing for a new major feature *prior* to it being comitted to CVS in a > > dev cycle than I think we ever gave any previous new feature even in the > > beta test phase. I don't object to testing, and have been thinking about > > coding something to address Tom's concerns, but let's demand heavy > > testing for the right reasons, not just to try to justify not doing a > > Win32 port. > > Nice try. I've demanded nothing, quit twisting my words to fit your > argument. If you're going to test and call it conclusive, do some > conclusive testing or call it something else. So we have no conclusive testing done that /proves/ postgres to be reliable ? I guess that such thing (positive conclusive reliability test) is impossible even in theory. But Dave has done some testing that could not prove the opposite and concluded that it is good enough for him. So I guess that his test were if fact "conclusive", if only just for him ;) Sometimes it is very hard to do the pull-the-plug test - I've seen people pondering over a HP server they could not switch off after accidentally powering it up. Pulling the plug just made it beep, but did not switch it off ;) > But I suspect that since > you don't know the difference between yanking the plug and kill -9 this > conversation is a waste of time. I assume you realize that U can't "kill -9" the plug ;) -- Hannu Krosing <hannu@tm.ee>
Hannu Krosing <hannu@tm.ee> writes:
> Kill -9 seems to me _less_ severe than yanking the plug but much easier
> to automate, so that could be the first thing to test. You have no hope
> of passing the pull-the-plug test if you can't survive even kill -9.
Actually, they're two orthogonal issues.
In the pull-the-plug case you have to worry about what is on disk at any
given instant and whether you can make all the bits on disk consistent
again. (And also about whether your filesystem can perform the
equivalent exercise for its own metadata; which is why we are
questioning Windows here. Oracle's Windows port may have an advantage,
if they bypass the OS to do raw disk I/O as they do on other platforms.)
In the kill -9 case there is no risk of losing data consistency on disk,
because the OS isn't crashing; whatever we last wrote we can expect to
read. The issue for kill -9 is whether we can deal with leftover
dynamic state, like pre-existing shared memory segments, pre-existing
SysV semaphores, TCP port numbers that the kernel won't reassign until
some timeout expires, that kind of fun stuff. The reason the TIP is
still there is that there are platforms on which that stuff doesn't work
very nicely. It's better to let the postmaster exit cleanly so that
that state gets cleaned up. I have no idea what the comparable issues
are for a native Windows port, but I bet there are some...
regards, tom lane
"Dave Page" <dpage@vale-housing.co.uk> writes:
> I would also point out that we already list the Cygwin port of
> PostgreSQL as supported. Who ever gave that the kind of testing people
> are demanding now? I think the worst case scenario will be that our
> Win32 port is far better than the existing 'supported' solution.
A good point --- but what this is really about is expectations. If we
support a native Windows port then people will probably think that it's
okay to run production databases on that setup; whereas I doubt many
people would think that about the Cygwin-based port. So what we need to
know is whether the platform is actually stable enough that that's a
reasonable thing to do; so that we can plaster the docs with appropriate
disclaimers if necessary. Windows, unlike the other OSes mentioned in
this thread, has a long enough and sorry enough track record that it
seems appropriate to run such tests ...
regards, tom lane
On Thursday 30 January 2003 11:12, Tom Lane wrote: > A good point --- but what this is really about is expectations. If we > support a native Windows port then people will probably think that it's > okay to run production databases on that setup; whereas I doubt many > people would think that about the Cygwin-based port. So what we need to > know is whether the platform is actually stable enough that that's a > reasonable thing to do; so that we can plaster the docs with appropriate > disclaimers if necessary. Windows, unlike the other OSes mentioned in > this thread, has a long enough and sorry enough track record that it > seems appropriate to run such tests ... I think it's just developer backlash to Win32. I am on record (see the archives) as not wanting the Win32 port -- but the vitriol I've seen in this thread from several people is entirely uncalled for and is sickening. Dave appears to have tested this Win32 beta at least as much as a regular PostgreSQL release would be tested. These tests are being held to artificially high standards, simply because it's native Win32. That is disgusting. And poor Katie just got _slammed_ -- and she's the lead developer. Vince, I would say that we, the developers of PostgreSQL, are then not qualified to test our own releases for the reasons you mentioned that Katie should not test her own releases. Of course that's ridiculous -- often the developers can do a better job of testing because they know better than the regular user would about what conditions can cause crashes. I don't like the thoughts of native Win32 either. I think Win32 should die a long horrible death. But that doesn't give me the right to publicly ridicule the folks that want to use PostgreSQL, even if it's in an 'industrial strength setting,' on Win32. The BSD license indemnifies us anyway. So what's the problem. The developers don't like Win32. That's the problem. But as to 'industrial strength testing' -- do ANY of our releases get this sort of testing on ANY platform? No, typically it's 'regression passed' 'Ok, it's supported on that platform.' -- Lamar Owen WGCR Internet Radio 1 Peter 4:11
Hi, On Thursday 30 January 2003 17:12, you wrote: > "Dave Page" <dpage@vale-housing.co.uk> writes: > > I would also point out that we already list the Cygwin port of > > PostgreSQL as supported. Who ever gave that the kind of testing people > > are demanding now? I think the worst case scenario will be that our > > Win32 port is far better than the existing 'supported' solution. > > A good point --- but what this is really about is expectations. If we > support a native Windows port then people will probably think that it's > okay to run production databases on that setup; whereas I doubt many > people would think that about the Cygwin-based port. So what we need to > know is whether the platform is actually stable enough that that's a > reasonable thing to do; so that we can plaster the docs with appropriate > disclaimers if necessary. Windows, unlike the other OSes mentioned in > this thread, has a long enough and sorry enough track record that it > seems appropriate to run such tests ... > > regards, tom lane > > ---------------------------(end of broadcast)--------------------------- > TIP 4: Don't 'kill -9' the postmaster Ah, well - I wanted to hold off but could not. First, a disclaimer: I don't like Windows at all. There, you got it. But: it's actually quite stable if you configure it well, and don't run the 3 million available 'dang, this looks nice' tools on it. Place it in the corner, let it run only server apps, and it serves well and stable. In my experience (and I have quite some experience in letting Win machines run in heavy-duty 24/7 production floors) they will happily run and not eat data until the some hardware breaks or disks overflow, just like any OS. So, please, don't let a 'I don't like it' kind of flamewar hinder a native port. And please no more 'not for production use' warnings - see above. Make this 'not for production use on workstations'. Greetings,Joerg -- Leading SW developer - S.E.A GmbH Mail: joerg.hessdoerfer@sea-gmbh.com WWW: http://www.sea-gmbh.com
On Thu, 30 Jan 2003, Lamar Owen wrote: > Vince, I would say that we, the developers of PostgreSQL, are then not > qualified to test our own releases for the reasons you mentioned that Katie > should not test her own releases. Of course that's ridiculous -- often the > developers can do a better job of testing because they know better than the > regular user would about what conditions can cause crashes. Don't twist what I said. My statement about Katie was that she has a knowledge of the port and the OS to the point where there are things that she knows are wrong to do and would avoid doing it. In the case of this port the idea is to make sure that those things that may cause the backend to close are something that SHOULD be tested. By their own admission they haven't been doing that. All they've done is loaded it down and made sure it continued to work. The other ports have a long history, the windows port has ZERO history. If you're being sickened now, how sick would you be if something went wrong and you started seeing things all over /. and other sites going on about how PG crashed and blew away some corporation's data and half the OS away on something that at worse should have only caused the backend to close? It won't matter that it was running on windows, it would have been a native port that was blessed by the PGDG. If anything, the resistance to this testing should sicken you. Vince. -- Fast, inexpensive internet service 56k and beyond! http://www.pop4.net/ http://www.meanstreamradio.com http://www.unknown-artists.com Internet radio: It's not file sharing, it's just radio.
Lamar Owen <lamar.owen@wgcr.org> writes:
> And poor Katie just got _slammed_ -- and she's the lead developer.
We could definitely do without the vitriol. I'd like to apologize if
anyone took anything I said as a personal attack. It wasn't meant that
way.
> The developers don't like Win32. That's the problem.
Sure, we're on record as not liking Windows. But:
> But as to 'industrial strength testing' -- do ANY of our releases get this
> sort of testing on ANY platform? No, typically it's 'regression passed' 'Ok,
> it's supported on that platform.'
Most variants of Unix are known to be pretty stable. Most variants of
Unix are known to follow the Unix standard semantics for sync() and
fsync(). I think we are entirely justified in doubting whether Windows
is a suitable platform for PG, and in wanting to run tests to find out.
Yes, we are holding Windows to a higher standard than we would for a
Unix variant.
Partly this is a matter of wanting to protect Postgres' reputation.
Just on sheer numbers, if there is a native Windows port then there are
likely to be huge numbers of people using Postgres on Windows. If
that's not going to be a reliable combination, we need to know it and
tell them so up-front. Otherwise, people will be blaming Postgres, not
Windows, when they lose data. It's an entirely different situation from
whether Postgres-on-Joe-Blow's-Unix-Variant loses data, first because of
visibility, and second because of the different user base. Am I being
paranoid to suspect that the average Postgres-on-Windows user will be
less clueful than the average Postgres-on-Unix user? I don't think so.
Between the population factors and Windows' hard-earned reputation for
unreliability, we would be irresponsible not to be asking tough
questions here. If the Windows partisans don't think Windows should be
held to a higher standard than the platforms we already deal with,
why not? Are they afraid that their platform won't pass the scrutiny?
regards, tom lane
On Thursday 30 January 2003 13:17, Vince Vielhaber wrote: > On Thu, 30 Jan 2003, Lamar Owen wrote: > > Vince, I would say that we, the developers of PostgreSQL, are then not > > qualified to test our own releases for the reasons you mentioned that > > Katie should not test her own releases. > Don't twist what I said. My statement about Katie was that she has a > knowledge of the port and the OS to the point where there are things > that she knows are wrong to do and would avoid doing it. Then she would not be honestly testing, would she? > admission they haven't been doing that. All they've done is loaded it > down and made sure it continued to work. The other ports have a long > history, the windows port has ZERO history. Do we do powerfail testing on a unix-type port now? That's not testing the port, incidentally, it's testing the OS, sync semantics aside. Do we hold the other ports to the same standards? Yes, the Win32 port is a substantial change from the Unix ports. Yes, it needs robust testing. But all the ports need that same grade of testing, not just Win32. And that type of testing is not being rigorously done on any port now, unless it is being done by a few that aren't announcing that they are doing it. And thanks to hardware write-back caching on many hard drives, powerfail testing may be moot regardless of OS or filesystem type. > If you're being sickened > now, how sick would you be if something went wrong and you started seeing > things all over /. and other sites going on about how PG crashed and > blew away some corporation's data and half the OS away on something > that at worse should have only caused the backend to close? Sick enough. But that applies to all our supported platforms, not just Win32. From what I've seen and heard the 'supported' Cygwin port will barf all over itself under high load. So, the first thing I personally would test for a Win32 native port is 'how well is it performing under load?' -- after it passes that I would then throw the more pathological cases at it. > It won't > matter that it was running on windows, it would have been a native > port that was blessed by the PGDG. So? How many users out there actually know about the PGDG? How many users have gotten PostgreSQL from their distributor of choice (whether a Linux distribution, the Cygwin distribution, FreeBSD ports, or wherever) and know nothing of PGDG or even postgresql.org? We make ourselves too important. I know enough to take all those sites with a shakerful of salt. But then again I know enough to know that the batboy didn't help Clinton or Bush do anything, 'Weekly World News' aside. We can't prevent the tabloid mentality regardless of what we do. Or don't do. The point being that if any release of anything labeled 'PostgreSQL', regardless of its status as blessed or not blessed (or even cursed) by the PGDG, does what you've said, PostgreSQL as a whole will suffer. Our blessing or cursing is meaningless to most users. Or, in slightly different words, if they can't be bothered to care that it's on Windows then they aren't going to care whether we gave it the Royal Seal of PGDG either. However, I'm sure the folks that are wanting to sell this Win32 native port care a whole lot about how much return business they get -- so I'm sure they care more about whether it is robustly tested than you give them credit. > If anything, the resistance to this testing should sicken you. There isn't any resistance to this testing that I've seen. ISTM that the resistance is to the idea of a 'supported' WIn32 native port. So, let's test the Win32 native beta using your scheme, and see what falls down. And let's test Linux, *BSD, HP-UX, and AIX using the same scheme and see if it falls down. Let's just be fair about the testing. The Win32 stuff is being proclaimed as beta already -- so none are being misled into thinking it's production grade right now. But it is passing those tests that hitherto have been thrown at it -- and it seems to be passing them well. -- Lamar Owen WGCR Internet Radio 1 Peter 4:11
On Thu, 30 Jan 2003, Lamar Owen wrote: > On Thursday 30 January 2003 13:17, Vince Vielhaber wrote: > > On Thu, 30 Jan 2003, Lamar Owen wrote: > > > Vince, I would say that we, the developers of PostgreSQL, are then not > > > qualified to test our own releases for the reasons you mentioned that > > > Katie should not test her own releases. > > > Don't twist what I said. My statement about Katie was that she has a > > knowledge of the port and the OS to the point where there are things > > that she knows are wrong to do and would avoid doing it. > > Then she would not be honestly testing, would she? She consider herself testing to her own standards as a windows user/ developer. Is that enough? IMO, No. I've been on both sides know that the windows user/developer doesn't hold things to the same standards as the unix user/developer. > > admission they haven't been doing that. All they've done is loaded it > > down and made sure it continued to work. The other ports have a long > > history, the windows port has ZERO history. > > Do we do powerfail testing on a unix-type port now? That's not testing the > port, incidentally, it's testing the OS, sync semantics aside. Do we hold > the other ports to the same standards? Yes, the Win32 port is a substantial > change from the Unix ports. Yes, it needs robust testing. But all the ports > need that same grade of testing, not just Win32. And that type of testing is > not being rigorously done on any port now, unless it is being done by a few > that aren't announcing that they are doing it. Since you're pretty much ignoring my reasoning, I'll give you the same consideration. The history of windows as a platform has shown itself to be rather fragile compared to unix. Before you respond to this, read Tom Lane's response and reply to that. Vince. -- Fast, inexpensive internet service 56k and beyond! http://www.pop4.net/ http://www.meanstreamradio.com http://www.unknown-artists.com Internet radio: It's not file sharing, it's just radio.
On Thursday 30 January 2003 16:54, Tom Lane wrote: > Lamar Owen <lamar.owen@wgcr.org> writes: > > And, by the way, who in their right mind tests a database server by > > repeated yanking of the AC power? > Anybody who would like their data to survive a power outage. I don't buy that. That's why I have $36,000 worth of lead acid in the room next door, with $5,000 of inverters and chargers in the server room. Until I had to upgrade RAM I had 240+ days of uptime on one box. The longest power interruption was 28 hours. The battery held the whole time. There was never more than 30 days between interruptions. The last time I had the server actually power down was during a maintenance run on the inverter/charge system, and I had to transfer power to the servers onto another branch, necessitating two power cycles, which were clean shutdown/reboots. I haven't had an unscheduled dirty powerdown in two years. We cannot on any system guarantee the data surviving a sudden power outage. Until we can be certain the write-back cache on that high performance drive (or NAS array using iSCSI, perhaps) flushes we cannot know the data hit the disks. > > To go to that extreme for Win32 when we caution > > against something as mundane as a kill -9 of postmaster on Unix is > > absurd. And, yes, I know the difference. I also know that the AC power > > pull has nothing to do with PostgreSQL, but it has to do with the OS > > under it. Although a kill -9, from the point of view of the running > > process, is identical to a power failure. > No, it is not. Did you not read my comments earlier today? Of course I did -- I'm not daft. And that's why I specified 'from the point of view of the running process' -- that is, the process you are SIGKILLing cannot itself determine the difference between the power cycle and SIGKILL. It just simply goes down, hard. Of course there is: > I forgot to mention one of the biggest > headaches, which is that kill -9 the postmaster doesn't kill the child > backends. This is a real difference, and one that I forgot as well. So SIGKILL is different to the whole backend system, but not to the singular process that is being SIGKILL'd. Suppose I issue a SIGKILL to postmaster and all forked backends simultaneously? Where does SIGKILL differ from a power failure from the point of view of the database system in that scenario? This is also assuming that you clean reboot the OS after the SIGKILL to postmaster, as there is that dynamic state you mentioned to worry about. I probably should have mentioned that before. > Windows > is going to bring a whole new set of failure modes that we don't have > defenses for. (Yet.) *That* is what we need extensive testing to learn > about, and claiming that we are discriminating against Windows just > because it's Windows misses the point completely. And ISTM that an experienced Windows developer, such as Katie or Dave, would know to do this, would know how to do this, and would know the best way of doing this. And I wasn't singling you out, Tom. It was the whole thread and the turns it took that got me rather upset. > Or, if you prefer, we can ship Postgres 7.4 for Windows with no more > testing than we need for any of the existing, long-since-well-tested > ports. But I'll bet a great deal that our reputation will go down the > drain (along with many people's data) if we do that. We don't have a standard testing methodology for any of our ports. We need one for all of our ports. I fully expect the Win32 port to need a different methodology than the FreeBSD port or the Linux port. And I expect we have enough experienced Win32 developers (which I am not) here that can provide insight into how the methodologies should differ. I prefer more extensive testing for all of our ports. You did read that when I wrote it, right? (When I wrote it multiple times....) Just saying 'it passed regression' shouldn't be enough -- but we should really spend some cycles thinking about what the test suite really should be. For all platforms. -- Lamar Owen WGCR Internet Radio 1 Peter 4:11
Lamar Owen <lamar.owen@wgcr.org> writes:
> And, by the way, who in their right mind tests a database server by repeated
> yanking of the AC power?
Anybody who would like their data to survive a power outage.
> To go to that extreme for Win32 when we caution
> against something as mundane as a kill -9 of postmaster on Unix is absurd.
> And, yes, I know the difference. I also know that the AC power pull has
> nothing to do with PostgreSQL, but it has to do with the OS under it.
> Although a kill -9, from the point of view of the running process, is
> identical to a power failure.
No, it is not. Did you not read my comments earlier today? The reasons
why we are concerned about kill -9 have *nothing* to do with whether the
database can survive system crashes. Rather, the issues created by kill
-9 have to do with coping with leftover state from a previous postmaster
in the same system lifecycle. I forgot to mention one of the biggest
headaches, which is that kill -9 the postmaster doesn't kill the child
backends. We've got an interlock that tries to prevent starting a new
postmaster when there are still old children around, but it's one of the
things that I think is most likely to break on any new port. (And I'm
dead certain that that code doesn't work on Windows.) It's that sort of
thing that we have painfully worked out on Unix-based systems, and are
going to have to do over again for Windows. In many places we are
probably not even going to realize that we have to do something over
again, until someone gets bitten.
The fact that Postgres is reliable does not come (only) from the code
being "right" in some abstract sense that will carry over to a new
platform. A big reason it's reliable is that we have painfully learned
about Unix-ish failure modes and put in defenses against them. Windows
is going to bring a whole new set of failure modes that we don't have
defenses for. (Yet.) *That* is what we need extensive testing to learn
about, and claiming that we are discriminating against Windows just
because it's Windows misses the point completely.
Or, if you prefer, we can ship Postgres 7.4 for Windows with no more
testing than we need for any of the existing, long-since-well-tested
ports. But I'll bet a great deal that our reputation will go down the
drain (along with many people's data) if we do that.
regards, tom lane
Lamar Owen <lamar.owen@wgcr.org> writes:
> While I understand (and agree with) your (and Vince's) reasoning on why
> Windows should be considered less reliable, neither of you have provided a
> sound technical basis for why we should not hold the other ports to the same
> standards.
The point here is that Windows is virgin territory for us. We know
about Unix. When we port to a new Unix variant, we are dealing with the
same system APIs, and in many cases large chunks of the same system
code, that we've dealt with before. It's reasonable for us to have
confidence that Postgres will work the same on such a platform as it
does on other Unix variants. And the track record of reliability that
we have built up across a bunch of Unix variants gives us
cross-pollinating confidence in all of them.
Windows shares none of that heritage. It is the first truly new port,
onto a system without any Unix background, that we have ever done AFAIK.
Claiming that it doesn't require an increased level of testing is
somewhere between ridiculous and irresponsible.
> I believe we should test every release as pathologically as Vince
> has stated for Win32.
Great, go to it. That does not alter the fact that today, with our
existing port history, Windows has to be treated with extra suspicion.
I do not buy the argument you are making that we should treat all
platforms alike. If we had a ten-year-old Windows port, we could
consider it as stable as all our other ten-year-old Unix ports.
We don't. Given that we don't have infinite resources for testing,
it's simple rationality to put more testing emphasis on the places
that we suspect there will be problems. And if you don't suspect
there will be problems on Windows, you are being way too naive :-(
> Do we want to encourage Win32? (some obviously do, but I don't) Well, telling
> people that we have tested PostgreSQL on Win32 much more thoroughly than on
> Unix is in a way telling them that we think it is _better_ than the
> time-tested Unix ports ('It passed a harder test on Win32. Are we afraid the
> Unix ports won't pass those same tests?').
If it passes the tests, good for it. I honestly do not expect that it
will. My take on this is that we want to be able to document the
problems in advance, rather than be blindsided.
regards, tom lane
Tom Lane wrote: > "Dave Page" <dpage@vale-housing.co.uk> writes: > > I would also point out that we already list the Cygwin port of > > PostgreSQL as supported. Who ever gave that the kind of testing people > > are demanding now? I think the worst case scenario will be that our > > Win32 port is far better than the existing 'supported' solution. > > A good point --- but what this is really about is expectations. If we > support a native Windows port then people will probably think that it's > okay to run production databases on that setup; Sure. But it's only common sense that a piece of software is only as reliable as the platform it's running on. People run production databases under MS-SQL all the time. Has MS-SQL itself gained a reputation for being an unreliable piece of junk? Perhaps. But if so, that obviously hasn't stopped people from putting their production databases on it! Is MS-SQL's reputation for unreliability, if any, because of MS-SQL itself or the platform it's operating on? The way to answer that is to ask the same question of Oracle and DB/2 under Windows. And therefore, the answer seems to be that the platform is a minor determinant, if any. > whereas I doubt many > people would think that about the Cygwin-based port. Why not? Seriously, if the people in question are the simpletons that you appear to be expecting them to be, then wouldn't they have that same expectation of the Cygwin based port? Why not? > So what we need to > know is whether the platform is actually stable enough that that's a > reasonable thing to do; so that we can plaster the docs with appropriate > disclaimers if necessary. Well, shouldn't we do that anyway, then, until we know otherwise? Shouldn't we do that with *any* new port? > Windows, unlike the other OSes mentioned in > this thread, has a long enough and sorry enough track record that it > seems appropriate to run such tests ... With this I agree, but before you start thinking that Windows is the only OS that qualifies, consider this: I've run the "pull the plug" test under early Linux 2.4 kernels running with ReiserFS. I'd start a make of a large project, pull the power, bring the system back up, and restart the build. And the end result was that some of the files files in the build directory were corrupted, such that the build could not continue. I haven't tried this under current versions of the kernel, so I don't know if things have improved or not. Doesn't that -- shouldn't that -- give you pause about declaring *Linux* an industrial-strength solution? My point: if you're going to hold *one* OS to a given standard, you should hold *all* of them to that same standard. -- Kevin Brown kevin@sysexperts.com
From: "Tom Lane" <tgl@sss.pgh.pa.us> > > Most variants of Unix are known to be pretty stable. Most variants of > Unix are known to follow the Unix standard semantics for sync() and > fsync(). I think we are entirely justified in doubting whether Windows > is a suitable platform for PG, and in wanting to run tests to find out. > Yes, we are holding Windows to a higher standard than we would for a > Unix variant. The patches that were released implement fsync() by a call to _commit(), which is what I expected to see after a brief tour of the M$ support site. Is there any reason to think this won't have the desired effect? IANAWD, but my reading suggests these should be pretty much equivalent. andrew
On Thu, Jan 30, 2003 at 02:39:59PM -0800, Kevin Brown wrote: > > With this I agree, but before you start thinking that Windows is the > only OS that qualifies, consider this: I've run the "pull the plug" > test under early Linux 2.4 kernels running with ReiserFS. I'd start a > make of a large project, pull the power, bring the system back up, and > restart the build. And the end result was that some of the files > files in the build directory were corrupted, such that the build could > not continue. Afaik, ReiserFS does not guarantee data consistency, only meta data. As in, the file system itself will be consistent, and an fsck shouldn't find a problem. Kurt
Kurt Roeckx wrote: > On Thu, Jan 30, 2003 at 02:39:59PM -0800, Kevin Brown wrote: > > > > With this I agree, but before you start thinking that Windows is the > > only OS that qualifies, consider this: I've run the "pull the plug" > > test under early Linux 2.4 kernels running with ReiserFS. I'd start a > > make of a large project, pull the power, bring the system back up, and > > restart the build. And the end result was that some of the files > > files in the build directory were corrupted, such that the build could > > not continue. > > Afaik, ReiserFS does not guarantee data consistency, only meta > data. As in, the file system itself will be consistent, and an > fsck shouldn't find a problem. Exactly. Does NTFS? Not as far as I know. Why should we hold NTFS to a standard that ReiserFS doesn't meet? That said, I do agree with Tom that the Windows port is basically virgin territory and needs to be approached with caution. But we shouldn't be so cautious that we hesitate to release the port to the world (sufficient disclaimers are appropriate, as with any new port)... -- Kevin Brown kevin@sysexperts.com
On Thursday 30 January 2003 15:29, Tom Lane wrote: > Lamar Owen <lamar.owen@wgcr.org> writes: > > While I understand (and agree with) your (and Vince's) reasoning on why > > Windows should be considered less reliable, neither of you have provided > Windows shares none of that heritage. It is the first truly new port, > onto a system without any Unix background, that we have ever done AFAIK. > Claiming that it doesn't require an increased level of testing is > somewhere between ridiculous and irresponsible. I am saying that as we mature we need increased testing across the board. And it is a very low percentage of code that is tied into the OS API, right? The majority of the code (the vast majority) isn't touched by it. > that we suspect there will be problems. And if you don't suspect > there will be problems on Windows, you are being way too naive :-( Reread my statement above. I _agree_ with the rationale -- but I fear it will have the opposite impact. And I am not convinced that just because we have good history with the unixoid ports means that we can slack on them -- Linux, *BSD, etc all change. The strftime(3) breakage with RedHat of a cycle ago should show us that much. I suspect there will be problems on Win32 -- it is, after all, a new port. But if we're going to immediately throw pathological test cases at it that we're not even bothering to test against now, that immediately throws up a flag to me. And TESTING IS BEING DONE on the Win32 port, nobody is yet trying to put the PGDG blessing on it as yet, and progress is being made by those who wish to see it made. It is still being touted as beta software, right? The patches from Jan are very preliminary still, correct? Katie hasn't issued a press release saying that it's not beta, right? <hyperbole> I don't see what the uproar is about, other than 'Win32 is so unstable that it can't possibly work as well as you are seeing it work -- you must be doing something wrong. Test it harder. Pull the plug repeatedly!! Test it until it breaks! HA! Told you it would break! (yeah, firing up the old oxyacetlyene torch and hitting the hard drive with a 6,000 degree flame did the trick -- this has got to be a bad operating system!)' </hyperbole> And, by the way, who in their right mind tests a database server by repeated yanking of the AC power? To go to that extreme for Win32 when we caution against something as mundane as a kill -9 of postmaster on Unix is absurd. And, yes, I know the difference. I also know that the AC power pull has nothing to do with PostgreSQL, but it has to do with the OS under it. Although a kill -9, from the point of view of the running process, is identical to a power failure. It simply dies (unless it becomes a zombie, in which case it is undead) either way. The effects of a kill -9 shouldn't be as severe as a power fail, since the OS can properly flush written buffers even after the process writing them has died. And I also can point the finger at some Unix swervers (spelling intentional) that would fail that test in a miserable way. I can also point at a few VMS machines that couldn't pass that test. I've even seen machines blow up due to improper power cycling. And I've seen Win2k machines come right up after repeated power blips (I've also seen them not come up). It really depends upon what the hard disk is doing at the instant the regulators drop out the 5 and 12V supplies (and which supply goes out first, which can depend upon the respective loads -- for modern Pentium 4 systems the 12V will probably go down first since it is more heavily loaded than the 5V supply in these systems). Under certain conditions where the 12V goes down before the 5V does, the head might still be writing as the servo spirals towards park, causing all manner of damage (maybe even to servo information, which normally cannot be written). So the power cycle becomes a test of hardware, too, played Russian Roulette-style. Talk about an unscientific test. A database server that needs that kind of testing is going to be hardened hardware on a doubly redundant UPS anyway. But, then again I've seen a Linux server survive a power cycle with no lost data (ext3 filesystem -- I've seen lost data with ext2). And I've seen the same server barf all over itself due to a single bit error in memory. Blew out the entire root filesystem, which was journaled and residing on a RAID 1 partition (the corruption was perfectly mirrored, by the way). Serves me right for not having ECC RAM installed at the time. > If it passes the tests, good for it. I honestly do not expect that it > will. My take on this is that we want to be able to document the > problems in advance, rather than be blindsided. I fully expect that Katie, Jan, Dave, and all the others working on this share your concerns and want the Win32 port to be as solid as is possible on that OS. -- Lamar Owen WGCR Internet Radio 1 Peter 4:11
On Thursday 30 January 2003 13:34, Tom Lane wrote:
> anyone took anything I said as a personal attack. It wasn't meant that
> way.
With a <flame on> tag? Flames are by long tradition personal. But I
understand that that wasn't the intent -- the <flame on> was more of a
<emphasis> tag.
> Sure, we're on record as not liking Windows. But:
> > But as to 'industrial strength testing' -- do ANY of our releases get
> > this sort of testing on ANY platform? No, typically it's 'regression
> > passed' 'Ok, it's supported on that platform.'
> Most variants of Unix are known to be pretty stable. Most variants of
> Unix are known to follow the Unix standard semantics for sync() and
> fsync(). I think we are entirely justified in doubting whether Windows
> is a suitable platform for PG, and in wanting to run tests to find out.
Testing is being done. Those who are testing it are comfortable so far in its
capabilities. We will hear about it, loadly, when that changes, I'm sure.
> Yes, we are holding Windows to a higher standard than we would for a
> Unix variant.
Which is pretty ironic, given Win's reputation, right?
> Partly this is a matter of wanting to protect Postgres' reputation.
And here's where the rubber meets the road. We, like many developers of
software (open source and otherwise) have worked on this for so long and so
hard that we have personified the program and it has become our child, so to
speak. As a father of four, I know what that can do. We will protect our
child at any cost, vehemently so. I for one can recognize this, and further
recognize that _it's_just_a_program_ (!!!!!) and not my child. This is hard
to do. We're seeing our child experiment with what we consider to be bad
company, and the defense mechanism is kicking in.
> Just on sheer numbers, if there is a native Windows port then there are
> likely to be huge numbers of people using Postgres on Windows. If
While I understand (and agree with) your (and Vince's) reasoning on why
Windows should be considered less reliable, neither of you have provided a
sound technical basis for why we should not hold the other ports to the same
standards. I believe we should test every release as pathologically as Vince
has stated for Win32. The more reliable we become, the worse our test cases
should become. Across the board, and not just on Win32.
Do we want to encourage Win32? (some obviously do, but I don't) Well, telling
people that we have tested PostgreSQL on Win32 much more thoroughly than on
Unix is in a way telling them that we think it is _better_ than the
time-tested Unix ports ('It passed a harder test on Win32. Are we afraid the
Unix ports won't pass those same tests?'). I for one don't want that to be a
conclusion -- but the 'suits' will see it that way, rest assured.
--
Lamar Owen
WGCR Internet Radio
1 Peter 4:11
Lamar Owen <lamar.owen@wgcr.org> writes:
> On Thursday 30 January 2003 16:54, Tom Lane wrote:
>> Lamar Owen <lamar.owen@wgcr.org> writes:
>>> And, by the way, who in their right mind tests a database server by
>>> repeated yanking of the AC power?
>> Anybody who would like their data to survive a power outage.
> I don't buy that. That's why I have $36,000 worth of lead acid in the room
> next door, with $5,000 of inverters and chargers in the server room.
Well, great; you're probably proof against misfeasance of your local
power company. But how about someone tripping over the power cord?
Or a blowout in the server's internal power supply? Or a kernel crash?
Pulling the power plug is just a convenient way of (approximately)
modeling a whole class of unpleasant events. I don't think the fact
that you can afford to spend that much on batteries makes it
uninteresting to test such scenarios.
But we're pretty much talking at cross-purposes here. The real issue
IMHO is that the Windows port needs a lot of testing because it is a
new platform (for us), and one not like the platforms we've used before.
It is faulty to equate the amount of testing required to gain confidence
in that port with the amount of testing required to gain confidence that
PG 7.4 will run reliably on, say, HPUX 10.20, when we already know that
every PG back to 6.4 has run reliably on HPUX 10.20. You're attacking a
straw man you have set up, namely the idea that only specific testing
produces confidence in a port. In my mind past track record has a lot
more to do with confidence than whatever testing we do for an individual
release.
regards, tom lane
Dave, Lamar and Katie can cheer now 'cuze this is the last comment I'm going to make on this. All others will be ignored, probably. The one thing I haven't seen from Dave, Lamar or Katie on this is reputation. You're all for the PostgreSQL name going on it but I have yet to see any of you so sure of yourselves that you'd put your own name on it. The license allows it. Red Hat did it. I see no "PageSQL" or "KatieSQL" or even an "Oh-Win SQL" being offered up. Yet all three of you are advocating that the PostgreSQL stamp of approval should be immediately placed on it (ok, Lamar may not be as in favor as the Dave and Katie). Without documented testing and sufficient warnings until enough history is banked, I don't think a native windows port should be given any kind of seal of approval. After that, what about keeping the code current? In a year or so will it suffer from bit-rot and be the source of complaints? Are there going to be security concerns surrounding it? Is there going to be a bunch of scrambling going on to put out a patch when the latest active-x bug hoses the data dir? Vince. -- Fast, inexpensive internet service 56k and beyond! http://www.pop4.net/ http://www.meanstreamradio.com http://www.unknown-artists.com Internet radio: It's not file sharing, it's just radio.
On Thursday 30 January 2003 18:39, Tom Lane wrote: > Well, great; you're probably proof against misfeasance of your local > power company. But how about someone tripping over the power cord? Twistlok. > Or a blowout in the server's internal power supply? Redundant supplies. > Or a kernel crash? Different from pulling the plug. > It is faulty to equate the amount of testing required to gain confidence > in that port with the amount of testing required to gain confidence that > PG 7.4 will run reliably on, say, HPUX 10.20, when we already know that > every PG back to 6.4 has run reliably on HPUX 10.20. But does the fact that PG 6.4 ran reliably on HP-UX 10 mean PG 7.4 will run as reliably on HP-UX 11? Does the fact that PG 6.2.1 ran well on Linux kernel 2.0.30 with libc 5.3.12 mean PG 7.4 will run well on Linux 2.6.x with glibc 2.4.x? The OS is also a moving target. Hmph. PG 7.3 won't even build on Red Hat 5.2, for instance. So much for track record. > You're attacking a > straw man you have set up, namely the idea that only specific testing > produces confidence in a port. In my mind past track record has a lot > more to do with confidence than whatever testing we do for an individual > release. Track record means nothing if sufficient items have changed in the underlying OS. I remember the Linux fiasco with PostgreSQL 6.3.1. It was so bad that Red Hat was considering releasing Red Hat 5.1 with a CVS checkout of pre-6.3.2. That is not Red Hat's normal policy. Also, between major versions enough may have changed to make it necessary to test thoroughly -- WAL, for instance. MVCC for another instance. PITR is going to be another instance requiring a different test methodology. One will indeed be required to blow down the whole system to properly test PITR, on all platforms. Track record indicates that all of our x.y.1 releases are typically hosed in some fashion. 7.3.1 proved that wrong. Track record only requires a single failure to invalidate -- and we should test for those failures across the board, regardless of track record. Records are meant to be broken. -- Lamar Owen WGCR Internet Radio 1 Peter 4:11
Tom Lane wrote: > Most variants of Unix are known to be pretty stable. Most variants of > Unix are known to follow the Unix standard semantics for sync() and > fsync(). I think we are entirely justified in doubting whether Windows > is a suitable platform for PG, and in wanting to run tests to find out. > Yes, we are holding Windows to a higher standard than we would for a > Unix variant. > > Partly this is a matter of wanting to protect Postgres' reputation. > Just on sheer numbers, if there is a native Windows port then there are > likely to be huge numbers of people using Postgres on Windows. If > that's not going to be a reliable combination, we need to know it and > tell them so up-front. Otherwise, people will be blaming Postgres, not > Windows, when they lose data. It's an entirely different situation from > whether Postgres-on-Joe-Blow's-Unix-Variant loses data, first because of > visibility, and second because of the different user base. Am I being > paranoid to suspect that the average Postgres-on-Windows user will be > less clueful than the average Postgres-on-Unix user? I don't think so. Assuming all your assumptions are right, why the hell is Oracle's and MS SQL-Server's reputation that bloody good? And what about MySQL? They all have a native Windows (sup)port for some time ... didn't harm their reputation. I think that we got in bed with this ugly Cybill ... er ... Cygwin thing had cost us more reputation than the sucking performance of pre-7 releases all together. Jan -- #======================================================================# # It's easier to get forgiveness for being wrong than for being right. # # Let's break this rule - forgive me. # #================================================== JanWieck@Yahoo.com #
On Thu, 30 Jan 2003, Dave Page wrote:
> > On Wed, 29 Jan 2003, Ron Mayer wrote:
> > >
> > > Cool irony in the automated .sig on the mailinglist software...
> > > [...]
> > > Sounds like you're basically saying is
> > > _do_ 'kill -9' the postmaster...
> > > and make sure it recovers gracefully...
> > ...
> It's not far off, but it's quite amusing none the less.
Sorry it looks like I should have added "50% :-), 50% :-|". I was
just amused by the irony of having the admonition against the
standard linux-low-memory condition. [ 90% :-) ]
More constructively, I think it'd be best for everyone if
a) postgresql does have a native windows port, and
b) it's positioned as a "well-tested beta" rather than "production ready", with documentation saying what part
peoplehave confidence in (the engine? operation in non-failure modes), and what part is still in beta (OS
failuremodes).
IMHO this would have the advantages of
Even if it works better than many expect, it shows... ... that postgresql has a high standard even for beta
releases, where this community thinks about even broader system issues.
... corporations using it what aspects they need to focus on in internal testing before using it in
productionenvironments.
... that we're interested in reaching a broader community.
... that postgresql is quite portable across platforms.
Note that I don't even care about running the windows version. I just think
that such a release can be positioned to _strengthen_ PostgreSQL's brand image
rather than weaken it.
Ron
Tom Lane wrote: > > Lamar Owen <lamar.owen@wgcr.org> writes: > > And, by the way, who in their right mind tests a database server by repeated > > yanking of the AC power? > > Anybody who would like their data to survive a power outage. ... has UPS, ECC Ram on quality boards and storage subsystems that guarantee the data to hit "some" surface after it passed the interface ... what's your point? Are you telling me that the reliability of an EMC2 system depends on which OS it is receiving the bits from? Is SuSE as reliable as TurboLinux? Or do I have to buy AIX to get the best result? Jan -- #======================================================================# # It's easier to get forgiveness for being wrong than for being right. # # Let's break this rule - forgive me. # #================================================== JanWieck@Yahoo.com #
Hannu Krosing wrote: > > I agree with Tom on yanking the plug while it's operating. Do you > > know the difference between kill -9 and yanking the plug? > > Kill -9 seems to me _less_ severe than yanking the plug but much easier > to automate, so that could be the first thing to test. You have no hope > of passing the pull-the-plug test if you can't survive even kill -9. > > Perhaps we could have a special "reliability-regression" test that does > "kill -9 postmaster", repeatedly, at random intervals, and checks for > consistency ? > > Maybe we will find even some options for some OS'es to "force-unmount" > disks. I guess that setting IDE disk's to read-only with hdparm could > possibly achieve something like that on Linux. Get VMWare for Linux, run whatever OS you like in it and "kill -9" the virtual machine. That's as close as you can get to "yanking" without wearing out your power plugs. Jan -- #======================================================================# # It's easier to get forgiveness for being wrong than for being right. # # Let's break this rule - forgive me. # #================================================== JanWieck@Yahoo.com #
Jan Wieck <JanWieck@Yahoo.com> writes:
> Assuming all your assumptions are right, why the hell is Oracle's and MS
> SQL-Server's reputation that bloody good?
They have marketing departments.
> And what about MySQL?
What about it? Someone claimed in this thread that MySQL's Windows port
requires Cygwin. Is that true or not?
regards, tom lane
> What about it? Someone claimed in this thread that MySQL's Windows port > requires Cygwin. Is that true or not? It's been a while, but I know I've installed MySQL on windows without any separate step of installing Cygwin (I can't say 100% for sure that it didn't install some part of Cygwin transparently to me). Regards,Jeff Davis
Jeff Davis wrote: >>What about it? Someone claimed in this thread that MySQL's Windows port >>requires Cygwin. Is that true or not? > > It's been a while, but I know I've installed MySQL on windows without any > separate step of installing Cygwin (I can't say 100% for sure that it didn't > install some part of Cygwin transparently to me). From the MySQL site's page about MySQL vs PostgreSQL: http://www.mysql.com/doc/en/MySQL-PostgreSQL_features.html "MySQL Server works better on Windows than PostgreSQL does. MySQL Server runs as a native Windows application (a service on NT/2000/XP), while PostgreSQL is run under the Cygwin emulation." That seems pretty straightforward. Regards and best wishes, Justin Clift > Regards, > Jeff Davis -- "My grandfather once told me that there are two kinds of people: those who work and those who take the credit. He told me to try to be in the first group; there was less competition there." - Indira Gandhi
On Thu, 2003-01-30 at 16:01, Vince Vielhaber wrote: > > Dave, Lamar and Katie can cheer now 'cuze this is the last comment > I'm going to make on this. All others will be ignored, probably. > > The one thing I haven't seen from Dave, Lamar or Katie on this is > reputation. You're all for the PostgreSQL name going on it but I > have yet to see any of you so sure of yourselves that you'd put > your own name on it. The license allows it. Red Hat did it. I > see no "PageSQL" or "KatieSQL" or even an "Oh-Win SQL" being offered > up. Yet all three of you are advocating that the PostgreSQL stamp > of approval should be immediately placed on it (ok, Lamar may not > be as in favor as the Dave and Katie). > Oh-win SQL! Man that was great :-) If only all of your posts were so witty... > Without documented testing and sufficient warnings until enough > history is banked, I don't think a native windows port should be > given any kind of seal of approval. After that, what about keeping > the code current? In a year or so will it suffer from bit-rot and > be the source of complaints? Are there going to be security concerns > surrounding it? Is there going to be a bunch of scrambling going on > to put out a patch when the latest active-x bug hoses the data dir? > We already support postgresql on cygwin, and we know that's crap. Having a native emulation can only improve that situation, so I don't see any reason not to move in that direction. All of this "stamp of approval" talk is really pointless at this juncture; no matter how much testing has been done, none of it means a lick until the code is integrated into the 7.4 branch. In the mean time, if some of the unix oriented guys want to devise a suggested test plan that can be used to determine if we are going to call the native windows support "production grade" or merely a vast improvement over the cygwin developers version, well I bet the windows folks would appreciate that. Even more so if someone runs those tests against a linux box so that we have actual statistics to compare against. Robert Treat
On Thu, 2003-01-30 at 20:29, Tom Lane wrote:
> Lamar Owen <lamar.owen@wgcr.org> writes:
> > While I understand (and agree with) your (and Vince's) reasoning on why
> > Windows should be considered less reliable, neither of you have provided a
> > sound technical basis for why we should not hold the other ports to the same
> > standards.
>
> The point here is that Windows is virgin territory for us. We know
> about Unix. When we port to a new Unix variant, we are dealing with the
> same system APIs, and in many cases large chunks of the same system
> code, that we've dealt with before. It's reasonable for us to have
> confidence that Postgres will work the same on such a platform as it
> does on other Unix variants. And the track record of reliability that
> we have built up across a bunch of Unix variants gives us
> cross-pollinating confidence in all of them.
>
> Windows shares none of that heritage. It is the first truly new port,
> onto a system without any Unix background, that we have ever done AFAIK.
I don't know how much Unix backgroun BeOS has. It does have a better
POSIX support than Win32, but I don't know how much of it is really from
Unix.
> Claiming that it doesn't require an increased level of testing is
> somewhere between ridiculous and irresponsible.
We should have at least _some_ platforms (besides Win32) that we could
clain to have run thorough test on.
I suspect that RedHat does some (perhaps even severe) testing for
RHAS/RHDB, but I don't know of any other thorough testing.
Or should reliability testing actually be something left for commercial
entities ?
> > I believe we should test every release as pathologically as Vince
> > has stated for Win32.
>
> Great, go to it. That does not alter the fact that today, with our
> existing port history, Windows has to be treated with extra suspicion.
I don't think that the pull-the-plug scenario happens enough in the wild
that even our seven-year track record can prove anything conlusive about
the reliability. I have not found instructions about providing that kind
of reliability in the docs either - things like what filesystems to use
on what OSes and with which mount options.
We just mention -f as a way to get non-reliable system ;)
> I do not buy the argument you are making that we should treat all
> platforms alike. If we had a ten-year-old Windows port, we could
> consider it as stable as all our other ten-year-old Unix ports.
> We don't. Given that we don't have infinite resources for testing,
> it's simple rationality to put more testing emphasis on the places
> that we suspect there will be problems. And if you don't suspect
> there will be problems on Windows, you are being way too naive :-(
"We" don't have that old windows port, but I guess that there are native
windows ports at least a few years old.
> > Do we want to encourage Win32? (some obviously do, but I don't) Well, telling
> > people that we have tested PostgreSQL on Win32 much more thoroughly than on
> > Unix is in a way telling them that we think it is _better_ than the
> > time-tested Unix ports ('It passed a harder test on Win32. Are we afraid the
> > Unix ports won't pass those same tests?').
>
> If it passes the tests, good for it. I honestly do not expect that it
> will. My take on this is that we want to be able to document the
> problems in advance, rather than be blindsided.
Where can I read such documentations for *nix ports ?
What I have read in this list is that losing different voltages in wrong
order can just write over any sectors on a disk, and that power-cycling
can blow up computers. I don't expect even Unix to survive that!
--
Hannu Krosing <hannu@tm.ee>
> Jeff Davis wrote: > >>What about it? Someone claimed in this thread that MySQL's Windows port > >>requires Cygwin. Is that true or not? > > > > It's been a while, but I know I've installed MySQL on windows without any > > separate step of installing Cygwin (I can't say 100% for sure that it didn' t > > install some part of Cygwin transparently to me). > > From the MySQL site's page about MySQL vs PostgreSQL: > http://www.mysql.com/doc/en/MySQL-PostgreSQL_features.html > > "MySQL Server works better on Windows than PostgreSQL does. MySQL Server > runs as a native Windows application (a service on NT/2000/XP), while > PostgreSQL is run under the Cygwin emulation." > > That seems pretty straightforward. But it's not /nearly/ that straightforward. If you look at the downloads that MySQL AB provides, they point you to a link that says "Windows binaries use the Cygwin library." Which apparently means that this "feature" is not actually a feature. Unlike PostgreSQL, which "is run under the Cygwin emulation," MySQL runs as a native Windows application (with Cygwin emulation). Apparently those are not at all the same thing, even though they are both using Cygwin... -- If this was helpful, <http://svcs.affero.net/rm.php?r=cbbrowne> rate me http://cbbrowne.com/info/linuxdistributions.html (1) Sigs are preceded by the "sigdashes" line, ie "\n-- \n" (dash-dash-space). (2) Sigs contain at least the name and address of the sender in the first line. (3) Sigs are at most four lines and at most eighty characters per line.
Jeff Davis wrote: > > What about it? Someone claimed in this thread that MySQL's Windows port > > requires Cygwin. Is that true or not? > > It's been a while, but I know I've installed MySQL on windows without any > separate step of installing Cygwin (I can't say 100% for sure that it didn't > install some part of Cygwin transparently to me). That may have involved "not being sufficiently observant," because the company quite clearly documents Cygwin as a dependancy. http://www.mysql.com/downloads/cygwin.html -- output = ("aa454" "@freenet.carleton.ca") http://www3.sympatico.ca/cbbrowne/linuxxian.html Change is inevitable, except from a vending machine.
> Jan Wieck <JanWieck@Yahoo.com> writes: > > Assuming all your assumptions are right, why the hell is Oracle's and MS > > SQL-Server's reputation that bloody good? > > They have marketing departments. ... As well as sizable systems integration departments devoted to the platforms in question. PostgreSQL doesn't have the latter, although the recent efforts make a move towards it. > > And what about MySQL? > > What about it? Someone claimed in this thread that MySQL's Windows port > requires Cygwin. Is that true or not? http://www.mysql.com/downloads/mysql-3.23.html "Windows downloads The Windows binaries use the Cygwin library. Source code for the version of Cygwin we have used is available on this page." http://www.mysql.com/downloads/cygwin.html -- (reverse (concatenate 'string "gro.gultn@" "enworbbc")) http://www.ntlug.org/~cbbrowne/spiritual.html "When you have eliminated the impossible, whatever remains, however improbable, must be the truth." -- Sir Arthur Conan Doyle (1859-1930), English author. Sherlock Holmes, in The Sign of Four, ch. 6 (1889). [...but see the Holmesian Fallacy, due to Bob Frankston... <http://www.frankston.com/public/Essays/Holmesian%20Fallacy.asp>]
On Thu, 2003-01-30 at 15:56, Tom Lane wrote: > The reason the TIP is > still there is that there are platforms on which that stuff doesn't work > very nicely. It's better to let the postmaster exit cleanly so that > that state gets cleaned up. I have no idea what the comparable issues > are for a native Windows port, but I bet there are some... That's why I proposed an automated test for this too. It is mostly important when conquering new OS'es, but could also be nice to have when testing if changes to storage manager or some other important subsystem will break anything. > regards, tom lane -- Hannu Krosing <hannu@tm.ee>
Christopher Browne wrote: <snip> >> From the MySQL site's page about MySQL vs PostgreSQL: >>http://www.mysql.com/doc/en/MySQL-PostgreSQL_features.html >> >>"MySQL Server works better on Windows than PostgreSQL does. MySQL Server >>runs as a native Windows application (a service on NT/2000/XP), while >>PostgreSQL is run under the Cygwin emulation." >> >>That seems pretty straightforward. > > But it's not /nearly/ that straightforward. > > If you look at the downloads that MySQL AB provides, they point you to a link > that says "Windows binaries use the Cygwin library." > > Which apparently means that this "feature" is not actually a feature. Unlike > PostgreSQL, which "is run under the Cygwin emulation," MySQL runs as a native > Windows application (with Cygwin emulation). Apparently those are not at all > the same thing, even though they are both using Cygwin... Hmm... wonder if they're meaning that MySQL compiles and executes as a True native windows application (skipping any unix compatibility calls), and it's just some of the support utils that use cygwin, or if they're trying to say that PostgreSQL has to operate entirely in the cygwin environment, whereas they don't? Regards and best wishes, Justin Clift -- "My grandfather once told me that there are two kinds of people: those who work and those who take the credit. He told me to try to be in the first group; there was less competition there." - Indira Gandhi
On Friday 31 January 2003 05:08, Tom Lane wrote: > Jan Wieck <JanWieck@Yahoo.com> writes: > > > And what about MySQL? > > What about it? Someone claimed in this thread that MySQL's Windows port > requires Cygwin. Is that true or not? For reference, from the INSTALL-SOURCE file included in the MySQL sources which I have lying about [*]: [*] danged legacy applications ;-) --QUOTE START-- Windows Source Distribution --------------------------- You will need the following: * VC++ 6.0 compiler (updated with 4 or 5 SP and Pre-processor package) The Pre-processor package is necessary for themacro assembler. More details at: `http://msdn.microsoft.com/vstudio/sp/vs6sp5/faq.asp'. * The MySQL source distribution for Windows, which can be downloaded from `http://www.mysql.com/downloads/'. Building MySQL 1. Create a work directory (e.g., workdir). 2. Unpack the source distribution in the aforementioned directory. 3. Start the VC++ 6.0 compiler. 4. In the `File' menu, select `Open Workspace'. 5. Open the `mysql.dsw' workspace you find on the work directory. 6. From the `Build' menu, select the `Set Active Configuration' menu. 7. Click over the screen selecting `mysqld - Win32 Debug' and click OK. 8. Press `F7' to begin the build of the debug server, libs, and some client applications. 9. When the compilation finishes, copy the libs and the executables to a separate directory. 10. Compile the release versions that you want, in the same way. 11. Create the directory for the MySQL stuff: e.g., `c:\mysql' 12. From the workdir directory copy for the c:\mysql directory the following directories: * Data * Docs * Share 13. Create the directory `c:\mysql\bin' and copy all the servers and clients that you compiled previously. 14. If you want, also create the `lib' directory and copy the libs that you compiled previously. 15. Do a clean using Visual Studio. Set up and start the server in the same way as for the binary Windows distribution. *Note Windows prepare environment::. --QUOTE END-- Ian Barwick barwick@gmx.net
On Fri, 2003-01-31 at 07:22, Christopher Browne wrote: > But it's not /nearly/ that straightforward. > > If you look at the downloads that MySQL AB provides, they point you to a link > that says "Windows binaries use the Cygwin library." > > Which apparently means that this "feature" is not actually a feature. Unlike > PostgreSQL, which "is run under the Cygwin emulation," MySQL runs as a native > Windows application (with Cygwin emulation). Apparently those are not at all > the same thing, even though they are both using Cygwin... I'm confused as to whether you are being sarcastic or truly seem to think there is a distinction here. Simple question, does MySQL require the cygwin dll's (or statically linked to) to run? If the answer is yes, then there is little question that they are as "emulated" as is the current PostgreSQL/Win32 effort. Care to expand on exactly what you believe the distinction is? ...or did I miss the humor boat? :( Regards, -- Greg Copeland <greg@copelandconsulting.net> Copeland Computer Consulting
Christopher Browne wrote: > <snip> > >> From the MySQL site's page about MySQL vs PostgreSQL: > >>http://www.mysql.com/doc/en/MySQL-PostgreSQL_features.html > >> > >>"MySQL Server works better on Windows than PostgreSQL does. MySQL > >>Server runs as a native Windows application (a service on > >>NT/2000/XP), while PostgreSQL is run under the Cygwin emulation." > >> > >>That seems pretty straightforward. > > > > But it's not /nearly/ that straightforward. > > > > If you look at the downloads that MySQL AB provides, they > point you to a link that says "Windows binaries use the Cygwin library." > > > > Which apparently means that this "feature" is not actually > a feature. > > Unlike > > PostgreSQL, which "is run under the Cygwin emulation," > MySQL runs as a native > > Windows application (with Cygwin emulation). Apparently > those are not at all > > the same thing, even though they are both using Cygwin... Justin Clift replied: > Hmm... wonder if they're meaning that MySQL compiles and > executes as a True native windows application (skipping any unix > compatibility calls), and it's just some of the support utils that > use cygwin, or if they're trying to say that PostgreSQL has to > operate entirely in the cygwin environment, whereas they don't? I just downloaded the latest productin source (3.3.55) and it appears to me that: 1) It uses Cygwin emulation via a dll. 2) It uses Visual Studio C++ 6.0 for the primary build environment. It compiles out of the box without having to learn Unix-style build systems, config, make, etc. No warnings, no errors, it just builds out of the box. If I did not have a lot of experience building databases I certainly would have found their support for Windows compelling. This is a big reason why they are #1. 3) The statement by the MySQL folks above that MySQL runs as a native Windows application (a service on NT/2000/XP) is indicative of why MySQL is kicking PostgreSQL's butt in terms of popularity. It is "marketing speak" at its best. It is technically true, MySQL runs as a service. As Christopher Browne points out, they still use the Cygwin Emulation layer. The statement is misleading, however, as it implies that they don't use any emulation but they do. The salient points: a) Running as a service is important as this the way NT/2000 administrators manage server tasks. The fact that PostgreSQL's Cygwin emulation doesn't do this is very indicative of inferior Windows support. b) MySQL recognizes that the important issue is to appear to be a well supported Windows application rather than to actually be one. c) It is probably much easier to add the support for running as an NT service than it is to write a true native port with no Cygwin dependency. NT Service support is basically a single funtion wrapper for certain API calls (startup, shutdown, etc.) that enable the Windows administration tools to deal with all servers in a similar manner. They have worked on that which makes them look better, makes their prospective customers happier, and makes it easier to support. Exactly what any good product development organization that listens to their customers would have done. <flame on> IMHO, PostgreSQL will never have the same level of use in the field as MySQL currently does as long as there is the kind "head in the sand" attitude about Windows that I've seen here on the hackers list, especially as evidenced by the recent outright attacks against those who are simply trying to port PostgreSQL to the largest platform out there today. There have been some very legitimate points about Windows being a new platform, one that will likely see a lot of users, and therefore one that should be more thoroughly tested before release than the typical port to another flavor of *nix. However, the way the conversation started reminds me of some of the chat discussions I've seen between young teens. I was a Mac developer way, way back and long ago realized that the best often loses and that better marketing beats better engineering every single time. <\flame off> DISCLAIMER: I hate Microsoft and Windows drives me nuts. - Curtis
----- Original Message ----- From: "Greg Copeland" <greg@CopelandConsulting.Net> > I'm confused as to whether you are being sarcastic or truly seem to > think there is a distinction here. Simple question, does MySQL require > the cygwin dll's (or statically linked to) to run? > > If the answer is yes, then there is little question that they are as > "emulated" as is the current PostgreSQL/Win32 effort. > > Care to expand on exactly what you believe the distinction is? ...or > did I miss the humor boat? :( I just installed it (their latest gama), to see what was there (and uninstalled it straight away ;-). There was a cygwinb19.dll (I think that's what it was called) installed. In any case, if we are talking about "industrial strength", is this the comparison we should be using? ;-) andrew
> On Fri, 2003-01-31 at 07:22, Christopher Browne wrote: > > But it's not /nearly/ that straightforward. >> If you look at the downloads that MySQL AB provides, they point you >> to a link that says "Windows binaries use the Cygwin library." >> Which apparently means that this "feature" is not actually a feature. >> Unlike PostgreSQL, which "is run under the Cygwin emulation," MySQL >> runs as a native Windows application (with Cygwin emulation). >> Apparently those are not at all the same thing, even though they are >> both using Cygwin... > I'm confused as to whether you are being sarcastic or truly seem to > think there is a distinction here. Simple question, does MySQL require > the cygwin dll's (or statically linked to) to run? I don't know if there's a distinction; read in whatever sarcasm is deserved by the reality of things. > If the answer is yes, then there is little question that they are as > "emulated" as is the current PostgreSQL/Win32 effort. Just so. If the answer is yes, then the MySQL folk are claiming an advantage that has no reality to it, in effect, "We aren't using Cygwin emulation, so we're better... (Whoops, we're actually /using/ Cygwin emulation.) > Care to expand on exactly what you believe the distinction is? ...or > did I miss the humor boat? :( I'm making the generous assumption that since /they/ claim that there is some distinction, that there perhaps is one. -- (concatenate 'string "cbbrowne" "@cbbrowne.com") http://cbbrowne.com/info/oses.html "All language designers are arrogant. Goes with the territory..." -- Larry Wall
Curtis Faith writes: > a) Running as a service is important as this the way NT/2000 > administrators manage server tasks. The fact that PostgreSQL's Cygwin > emulation doesn't do this is very indicative of inferior Windows > support. No, it is indicative of the inability to read the documentation. PostgreSQL on Cygwin runs as a service if and only if you ask it to. -- Peter Eisentraut peter_e@gmx.net
On Fri, 2003-01-31 at 16:07, Christopher Browne wrote: > I'm making the generous assumption that since /they/ claim that there is > some distinction, that there perhaps is one. I've used the cygwin environment enough to know that there isn't any. If it's linked against the cygwin dll, the application runs in an "emulated unix environment." To say it's emulated is really too strong but to say it adds *tons* of overhead certainly won't make you a lair. ;) -- Greg Copeland <greg@copelandconsulting.net> Copeland Computer Consulting
On Sat, 1 Feb 2003, Peter Eisentraut wrote: > Curtis Faith writes: > > > a) Running as a service is important as this the way NT/2000 > > administrators manage server tasks. The fact that PostgreSQL's Cygwin > > emulation doesn't do this is very indicative of inferior Windows > > support. > > No, it is indicative of the inability to read the documentation. > PostgreSQL on Cygwin runs as a service if and only if you ask it to. I would say that not supporting those who have an inability to read documentation would count as "inferior Windows support." :-) What I'm hearing here is that all we really need to do to "compete" with MySQL on Windows is to make the UI a bit slicker. So what's the problem with someone building, for each release, a set of appropriate binaries, and someone making a slick install program that will install postgres, install parts of cygwin if necessary, and set up postgres as a service? cjs -- Curt Sampson <cjs@cynic.net> +81 90 7737 2974 http://www.netbsd.org Don't you know, in this new Dark Age, we're alllight. --XTC
Curt Sampson wrote: <snip>> What I'm hearing here is that all we really need to do to "compete" with> MySQL on Windows is to make the UI a bitslicker. So what's the problem> with someone building, for each release, a set of appropriate binaries, and> someone making a slick install program that will install postgres,> install parts of cygwin if necessary,and set up postgres as a service? The non-code related parts of the Win32 port of PostgreSQL that are being looked at: + Working on the packaging bits (slick install program) already. Have created a project - pgsqlwin - on GBorg to hold any specific bits we need. http://gborg.postgresql.org/project/pgsqlwin/projdisplay.php First release of the *extremely alpha* "Proof of Concept" version is at: http://prdownloads.sourceforge.net/pgsql/PgSQL731wina1.exe?download + Concerned about including GPL stuff without having 100% totally investigated the ramifications for people including the Win32 version of PostgreSQL as a built-in part of their applications. Not going to commit anything even slightly GPL related to that GBorg project until it 100% safe to do so without affect our ability to release it as BSD. Have some preliminary information regarding this, but just need to wrap my head around it properly. Not going to look at it closely for another week or so. + It would be greatly helpful to have some way for the install program to automatically add the "Log in as a service" Win32 priviledge to the "postgres" user without having to instruct the user to do so. We can create the user automatically through a shell command, but no idea how to add that permission. If someone could do some Win32 API stuff to do it behind the scenes without a shell command even, that would be great. + The WinMaster project is a first go at creating a Win32 GUI command console for controlling the PostgreSQL service. It's still a bit too basic for real use though: http://gborg.postgresql.org/project/winmaster/projdisplay.php Further suggestions, volunteers, etc are totally welcome. :-) Regards and best wishes, Justin Clift > cjs -- "My grandfather once told me that there are two kinds of people: those who work and those who take the credit. He told me to try to be in the first group; there was less competition there." - Indira Gandhi
----- Original Message ----- From: "Justin Clift" <justin@postgresql.org> To: "Curt Sampson" <cjs@cynic.net> Cc: "Peter Eisentraut" <peter_e@gmx.net>; "Curtis Faith" <curtis@galtcapital.com>; <pgsql-hackers@postgresql.org> Sent: Sunday, February 02, 2003 4:42 AM Subject: Re: [mail] Re: [HACKERS] Windows Build System > + It would be greatly helpful to have some way for the install program > to automatically add the "Log in as a service" Win32 priviledge to the > "postgres" user without having to instruct the user to do so. We can > create the user automatically through a shell command, but no idea how > to add that permission. If someone could do some Win32 API stuff to do > it behind the scenes without a shell command even, that would be great. > > + The WinMaster project is a first go at creating a Win32 GUI command > console for controlling the PostgreSQL service. It's still a bit too > basic for real use though: > http://gborg.postgresql.org/project/winmaster/projdisplay.php> > Further suggestions, volunteers, etc are totally welcome. > :-) > Regards and best wishes, > Justin Clift It's still a bit too basic for real use though: Yeah i know. I write this for my internal use. Initial purpose of this stuff is only to avoid teaching of an old lady with minimum computer skills to use bash and hide this ugly dos box :) Mark L. Woodward (mlw) anounce few monts ago a self installing PostgreSQL for Windows so i write him about this console. He do a lof job to. Special thanks Mark. OK, now how to make WinMaster more usefull ? It's open source so if any1 want use it he/she may help to develop it. I. Install as a service feature for winmaster are included in my plans for future. II. I'm thinking about direct link to PostgreSQL server instead usung CreateProcess, but this is unclear idea at present time. Any suggestions will be welcome. III Please add any feature rquests to http://gborg.postgresql.org/project/winmaster/bugs/buglist.php?fr=yes and ideas to mailto:winmaster-general@gborg.postgresql.org Justin you are right !!! Further suggestions, volunteers, etc are totally welcome!!! Further suggestions, volunteers, etc are totally welcome!!! Further suggestions, volunteers, etc are totally welcome!!! --- Outgoing mail is certified Virus Free. Checked by AVG anti-virus system (http://www.grisoft.com). Version: 6.0.449 / Virus Database: 251 - Release Date: 27/01/2003
Hannu Krosing wrote: > > On Thu, 2003-01-30 at 20:29, Tom Lane wrote: > > Claiming that it doesn't require an increased level of testing is > > somewhere between ridiculous and irresponsible. > > We should have at least _some_ platforms (besides Win32) that we could > clain to have run thorough test on. > > I suspect that RedHat does some (perhaps even severe) testing for > RHAS/RHDB, but I don't know of any other thorough testing. > > Or should reliability testing actually be something left for commercial > entities ? The testing has to be done before we make anything available as an official release. As of now, the status of this project is at the beginning of incorporating a 7.2.1 based patch into CVS HEAD. Asking for exzessive tests at this stage of development and (ab)using the absence of 100% proof of rock solid reliability as an excuse to reject the entire aproach would be ridiculous. Jan -- #======================================================================# # It's easier to get forgiveness for being wrong than for being right. # # Let's break this rule - forgive me. # #================================================== JanWieck@Yahoo.com #