Thread: archives.postgresql.org change
Hi, We've just changed the archives site so that we only pull the mboxes from hub.org, and mhonarc is run locally on archives.pg.org. This is quicker -- we're now regenerating the archives every 15 minutes. Let me know (via this list, I guess) if anything weird crops up. -- Alvaro Herrera http://www.CommandPrompt.com/ The PostgreSQL Company - Command Prompt, Inc.
On Fri, 2008-02-22 at 18:12 -0300, Alvaro Herrera wrote: > Hi, > > We've just changed the archives site so that we only pull the mboxes > from hub.org, and mhonarc is run locally on archives.pg.org. This is > quicker -- we're now regenerating the archives every 15 minutes. > > Let me know (via this list, I guess) if anything weird crops up. Why was this done without any warning to the -www or sysadmins team? Aside from that, I assume you have verified that the search system still pulls things in properly? That's where I'd see a potential issue, but I don't have time to look into it right now (need to fix broken slides for tomorrows FOSDEM talks) Oh, and since you've been fiddling it, please get some documentation up on pmt. We didn't have any for the old one that I can find, but we'd really like some for the new one. (for one, we've been talking here at FOSDEM about auto-syncing some things about the lists between that one and the search stuff so we stop missing lists when they are added - and it'd be nice to have a documented system to work off for that one) //Magnus
Magnus Hagander wrote: > On Fri, 2008-02-22 at 18:12 -0300, Alvaro Herrera wrote: >> Hi, >> >> We've just changed the archives site so that we only pull the mboxes >> from hub.org, and mhonarc is run locally on archives.pg.org. This is >> quicker -- we're now regenerating the archives every 15 minutes. >> >> Let me know (via this list, I guess) if anything weird crops up. > > Why was this done without any warning to the -www or sysadmins team? Uhh... this has been being talking about for *months*. > > Aside from that, I assume you have verified that the search system still > pulls things in properly? That's where I'd see a potential issue, but I > don't have time to look into it right now (need to fix broken slides for > tomorrows FOSDEM talks) I think you are missing what we did :) It used to be that we rsynced all of archives.postgresql.org (website, php files, mbox files etc..). Now we just rsync mbox. Basically we are saving bandwidth and allowing marc to deccomission and old fbsd4 box. Operationally for all dependent stuff nothing should change. The locations, urls, ips etc.. didn't change. > > Oh, and since you've been fiddling it, please get some documentation up > on pmt. We didn't have any for the old one that I can find, but we'd > really like some for the new one. (for one, we've been talking here at > FOSDEM about auto-syncing some things about the lists between that one > and the search stuff so we stop missing lists when they are added - and > it'd be nice to have a documented system to work off for that one) Sure. Joshua D. Drake
On Sat, 2008-02-23 at 11:25 -0800, Joshua D. Drake wrote: > Magnus Hagander wrote: > > On Fri, 2008-02-22 at 18:12 -0300, Alvaro Herrera wrote: > >> Hi, > >> > >> We've just changed the archives site so that we only pull the mboxes > >> from hub.org, and mhonarc is run locally on archives.pg.org. This is > >> quicker -- we're now regenerating the archives every 15 minutes. > >> > >> Let me know (via this list, I guess) if anything weird crops up. > > > > Why was this done without any warning to the -www or sysadmins team? > > Uhh... this has been being talking about for *months*. Really? I've missed that completely. > > Aside from that, I assume you have verified that the search system still > > pulls things in properly? That's where I'd see a potential issue, but I > > don't have time to look into it right now (need to fix broken slides for > > tomorrows FOSDEM talks) > > I think you are missing what we did :) > > It used to be that we rsynced all of archives.postgresql.org (website, > php files, mbox files etc..). > > Now we just rsync mbox. Basically we are saving bandwidth and allowing > marc to deccomission and old fbsd4 box. Operationally for all dependent > stuff nothing should change. The locations, urls, ips etc.. didn't change. I understand perfectly well what you did. But we've seen *a lot* of cases before when moving one of these platform independent programs (like mhonarc, which is written in perl IIRC) between Linux and FreeBSD and get significantly different behaviour. Either in that it crashes, or more often in that it gives different results without crashing. What you did was move one of our core services from FreeBSD to Linux, which potentially has a lot of effects. I'm not saying that something broke, I'm just asking how much you actually verified. AFAIK, we also had a bunch of customizations in mhonarc - did those get migrated over, or are we now running a stock mhonarc? > > Oh, and since you've been fiddling it, please get some documentation up > > on pmt. We didn't have any for the old one that I can find, but we'd > > really like some for the new one. (for one, we've been talking here at > > FOSDEM about auto-syncing some things about the lists between that one > > and the search stuff so we stop missing lists when they are added - and > > it'd be nice to have a documented system to work off for that one) > > Sure. Good, thanks! Oh, and I also notice there has been zero commits about this to the autobackup repository, so obviously none of this is included in said backups - that also needs to be fixed. //Magnus
On Sun, Feb 24, 2008 at 9:53 AM, Magnus Hagander <magnus@hagander.net> wrote: > > > Uhh... this has been being talking about for *months*. > > Really? I've missed that completely. Yeah, me too. I know we were aiming to upgrade svr5, but don't recall any chatter about moving the archives generation. Marc normally just moves VMs to a FreeBSD 6 host and upgrades them. > I understand perfectly well what you did. But we've seen *a lot* of > cases before when moving one of these platform independent programs > (like mhonarc, which is written in perl IIRC) between Linux and FreeBSD > and get significantly different behaviour. Yeah - case in point - the new archives system I've been working on. Works fine on Linux, on FreeBSD PHP core dumps and occasionally throws mysterious out of memory errors and ignores the remaining messages in an mbox. > AFAIK, we also had a bunch of customizations in mhonarc - did those get > migrated over, or are we now running a stock mhonarc? Yeah, I'm not sure what they all were, but I do remember Marc spending a bunch of time hacking stuff around on Oleg & John's advice to optimise the way Google & other search engine index the archives. I don't know if all that customisation is stuff that can be rsync'd. -- Dave Page EnterpriseDB UK: http://www.enterprisedb.com The Oracle-compatible database company
On Sun, Feb 24, 2008 at 09:53:10AM +0100, Magnus Hagander wrote: > > > > Uhh... this has been being talking about for *months*. > > Really? I've missed that completely. I also do not remember any posting of an announcement like, "We're moving service $X from $machineA to $machineB at $timestamptz." There seems to be a predilection for moving things around without a lot of clear planning. If we're going to have a group of administrators maintaining machines, surely we need to have clear announcements of the above sort before and after such moves, so that nobody is surprised, no? A
Magnus Hagander wrote: > On Sat, 2008-02-23 at 11:25 -0800, Joshua D. Drake wrote: >> Magnus Hagander wrote: >>> On Fri, 2008-02-22 at 18:12 -0300, Alvaro Herrera wrote: >>>> Hi, > AFAIK, we also had a bunch of customizations in mhonarc - did those get > migrated over, or are we now running a stock mhonarc? > We only did what Marc told us we had to do. We had a couple of scripts that he provided that we had to port from tcsh (yes) to sh, which Alvaro kindly did. As I understand it all the scripts did was format the messages from the mbox files but to be honest, I didn't read them. I had Alvaro and Marc work it out (I refuse to touch tcsh). > >>> Oh, and since you've been fiddling it, please get some documentation up >>> on pmt. We didn't have any for the old one that I can find, but we'd >>> really like some for the new one. (for one, we've been talking here at >>> FOSDEM about auto-syncing some things about the lists between that one >>> and the search stuff so we stop missing lists when they are added - and >>> it'd be nice to have a documented system to work off for that one) >> Sure. > > Good, thanks! > > Oh, and I also notice there has been zero commits about this to the > autobackup repository, so obviously none of this is included in said > backups - that also needs to be fixed. Its on the list :) Joshua D. Drake
On Sun, 2008-02-24 at 07:53 -0800, Joshua D. Drake wrote: > Magnus Hagander wrote: > > On Sat, 2008-02-23 at 11:25 -0800, Joshua D. Drake wrote: > >> Magnus Hagander wrote: > >>> On Fri, 2008-02-22 at 18:12 -0300, Alvaro Herrera wrote: > >>>> Hi, > > > AFAIK, we also had a bunch of customizations in mhonarc - did those get > > migrated over, or are we now running a stock mhonarc? > > > > We only did what Marc told us we had to do. We had a couple of scripts > that he provided that we had to port from tcsh (yes) to sh, which Alvaro > kindly did. As I understand it all the scripts did was format the > messages from the mbox files but to be honest, I didn't read them. I had > Alvaro and Marc work it out (I refuse to touch tcsh). As long as these things are covered, that's fine. But it'd be good if someone could verify it who actually knew about it. I think that means you, Marc :-) But if that's already done, then drop the point - it just wasn't clear from the original message. > >>> Oh, and since you've been fiddling it, please get some documentation up > >>> on pmt. We didn't have any for the old one that I can find, but we'd > >>> really like some for the new one. (for one, we've been talking here at > >>> FOSDEM about auto-syncing some things about the lists between that one > >>> and the search stuff so we stop missing lists when they are added - and > >>> it'd be nice to have a documented system to work off for that one) > >> Sure. > > > > Good, thanks! > > > > Oh, and I also notice there has been zero commits about this to the > > autobackup repository, so obviously none of this is included in said > > backups - that also needs to be fixed. > > Its on the list :) Yay :-) //Magnus
Magnus Hagander wrote: > As long as these things are covered, that's fine. But it'd be good if > someone could verify it who actually knew about it. I think that means > you, Marc :-) But if that's already done, then drop the point - it just > wasn't clear from the original message. And Marc has been in the loop the whole time (just fyi). >>> >>> Oh, and I also notice there has been zero commits about this to the >>> autobackup repository, so obviously none of this is included in said >>> backups - that also needs to be fixed. Yeah... that has just been done. Joshua D. Drake
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 - --On Sunday, February 24, 2008 10:42:45 +0100 Dave Page <dpage@pgadmin.org> wrote: > Yeah, me too. I know we were aiming to upgrade svr5, but don't recall > any chatter about moving the archives generation. Marc normally just > moves VMs to a FreeBSD 6 host and upgrades them. Actually, I had asked Joshua to do this months ago ... figured it made more sense to pull justhte mbox files to the archives server and regen, then pull them to an intermediary box, and then have archives pull the generated files ... the middle step didn't make much sense ... > Yeah, I'm not sure what they all were, but I do remember Marc spending > a bunch of time hacking stuff around on Oleg & John's advice to > optimise the way Google & other search engine index the archives. I > don't know if all that customisation is stuff that can be rsync'd. The only thing that that should have affected would have been the .resource file ... Joshua, we did copy that over, right? - ---- Marc G. Fournier Hub.Org Networking Services (http://www.hub.org) Email . scrappy@hub.org MSN . scrappy@hub.org Yahoo . yscrappy Skype: hub.org ICQ . 7615664 -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.4 (FreeBSD) iD8DBQFHwbZk4QvfyHIvDvMRAiAJAKCrGsBnk5TRtBPw4hcRdtbowQQdbQCgyu5i 9cPAoBYqWSB/WZUZk7T3Y+Q= =0tij -----END PGP SIGNATURE-----
On Sun, 24 Feb 2008 14:24:36 -0400
"Marc G. Fournier" <scrappy@hub.org> wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
>
>
> - --On Sunday, February 24, 2008 10:42:45 +0100 Dave Page
> <dpage@pgadmin.org> wrote:
>
>
> > Yeah, me too. I know we were aiming to upgrade svr5, but don't
> > recall any chatter about moving the archives generation. Marc
> > normally just moves VMs to a FreeBSD 6 host and upgrades them.
>
> Actually, I had asked Joshua to do this months ago ... figured it
> made more sense to pull justhte mbox files to the archives server and
> regen, then pull them to an intermediary box, and then have archives
> pull the generated files ... the middle step didn't make much
> sense ...
>
> > Yeah, I'm not sure what they all were, but I do remember Marc
> > spending a bunch of time hacking stuff around on Oleg & John's
> > advice to optimise the way Google & other search engine index the
> > archives. I don't know if all that customisation is stuff that can
> > be rsync'd.
>
> The only thing that that should have affected would have been
> the .resource file ... Joshua, we did copy that over, right?
I have the following:
-rw-r--r-- 1 archives archives 8041 Jul 13 2006 .resource-new
-rw-r--r-- 1 archives archives 7856 Aug 9 2006 .resource-old
Which contain a bunch of mhonarc macros (at least that is what it looks
like). It is called from mk-mhonarc like so:
if [ `echo $month | awk -F. '{print $2}'` -le 200607 ]; then resourcefile=.resource-old else
resourcefile=.resource-new fi
Sincerely,
Joshua D. Drake
--
The PostgreSQL Company since 1997: http://www.commandprompt.com/
PostgreSQL Community Conference: http://www.postgresqlconference.org/
Donate to the PostgreSQL Project: http://www.postgresql.org/about/donate
PostgreSQL SPI Liaison | SPI Director | PostgreSQL political pundit
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Perfect, just wanted to make sure ...
- --On Sunday, February 24, 2008 10:35:54 -0800 "Joshua D. Drake"
<jd@commandprompt.com> wrote:
> On Sun, 24 Feb 2008 14:24:36 -0400
> "Marc G. Fournier" <scrappy@hub.org> wrote:
>
>> -----BEGIN PGP SIGNED MESSAGE-----
>> Hash: SHA1
>>
>>
>>
>> - --On Sunday, February 24, 2008 10:42:45 +0100 Dave Page
>> <dpage@pgadmin.org> wrote:
>>
>>
>> > Yeah, me too. I know we were aiming to upgrade svr5, but don't
>> > recall any chatter about moving the archives generation. Marc
>> > normally just moves VMs to a FreeBSD 6 host and upgrades them.
>>
>> Actually, I had asked Joshua to do this months ago ... figured it
>> made more sense to pull justhte mbox files to the archives server and
>> regen, then pull them to an intermediary box, and then have archives
>> pull the generated files ... the middle step didn't make much
>> sense ...
>>
>> > Yeah, I'm not sure what they all were, but I do remember Marc
>> > spending a bunch of time hacking stuff around on Oleg & John's
>> > advice to optimise the way Google & other search engine index the
>> > archives. I don't know if all that customisation is stuff that can
>> > be rsync'd.
>>
>> The only thing that that should have affected would have been
>> the .resource file ... Joshua, we did copy that over, right?
>
> I have the following:
>
> -rw-r--r-- 1 archives archives 8041 Jul 13 2006 .resource-new
> -rw-r--r-- 1 archives archives 7856 Aug 9 2006 .resource-old
>
> Which contain a bunch of mhonarc macros (at least that is what it looks
> like). It is called from mk-mhonarc like so:
>
> if [ `echo $month | awk -F. '{print $2}'` -le 200607 ]; then
> resourcefile=.resource-old
> else
> resourcefile=.resource-new
> fi
>
> Sincerely,
>
> Joshua D. Drake
>
> --
> The PostgreSQL Company since 1997: http://www.commandprompt.com/
> PostgreSQL Community Conference: http://www.postgresqlconference.org/
> Donate to the PostgreSQL Project: http://www.postgresql.org/about/donate
> PostgreSQL SPI Liaison | SPI Director | PostgreSQL political pundit
>
- ----
Marc G. Fournier Hub.Org Networking Services (http://www.hub.org)
Email . scrappy@hub.org MSN . scrappy@hub.org
Yahoo . yscrappy Skype: hub.org ICQ . 7615664
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.4 (FreeBSD)
iD8DBQFHwbpS4QvfyHIvDvMRAhFYAJ0UrYH3jF3MdE4AgfgsYvSRsI03uQCcC4wa
MyN85lOf3QWhVfXcfCOMJWs=
=8BqQ
-----END PGP SIGNATURE-----
Magnus Hagander wrote: > > On Fri, 2008-02-22 at 18:12 -0300, Alvaro Herrera wrote: > > Hi, > > > > We've just changed the archives site so that we only pull the mboxes > > from hub.org, and mhonarc is run locally on archives.pg.org. This is > > quicker -- we're now regenerating the archives every 15 minutes. > > > > Let me know (via this list, I guess) if anything weird crops up. > > Why was this done without any warning to the -www or sysadmins team? Sorry, I don't know the answer to that. I kept looking for you, Dave or Stefan on Jabber, but saw none. > Aside from that, I assume you have verified that the search system still > pulls things in properly? That's where I'd see a potential issue, but I > don't have time to look into it right now (need to fix broken slides for > tomorrows FOSDEM talks) Good question -- I didn't. > Oh, and since you've been fiddling it, please get some documentation up > on pmt. If you get me a password for accessing pmt, I'll gladly document the whole shebang (which really needs some documenting -- the whole setup is not entirely trivial). Also, I'd like to put the scripts somewhere in a community SCM (I put them in a CMD SVN as a temporary measure.) > We didn't have any for the old one that I can find, but we'd > really like some for the new one. (for one, we've been talking here at > FOSDEM about auto-syncing some things about the lists between that one > and the search stuff so we stop missing lists when they are added - and > it'd be nice to have a documented system to work off for that one) That would be nice -- also the current scripts to generate the list of lists contain some duplicate info; it's no wonder that the lists has been out of date continuously. I can lend a hand with that as needed. -- Alvaro Herrera http://www.CommandPrompt.com/ PostgreSQL Replication, Consulting, Custom Development, 24x7 support
Alvaro Herrera wrote: > Magnus Hagander wrote: > > Aside from that, I assume you have verified that the search system still > > pulls things in properly? That's where I'd see a potential issue, but I > > don't have time to look into it right now (need to fix broken slides for > > tomorrows FOSDEM talks) > > Good question -- I didn't. Ok, it seems to work. I searched for "archives.postgresql.org change" and this thread shows up. -- Alvaro Herrera http://www.CommandPrompt.com/ The PostgreSQL Company - Command Prompt, Inc.
Alvaro Herrera wrote: > Alvaro Herrera wrote: >> Magnus Hagander wrote: > >>> Aside from that, I assume you have verified that the search system still >>> pulls things in properly? That's where I'd see a potential issue, but I >>> don't have time to look into it right now (need to fix broken slides for >>> tomorrows FOSDEM talks) >> Good question -- I didn't. > > Ok, it seems to work. I searched for "archives.postgresql.org change" > and this thread shows up. > A note on the svn, the files are now in the autobackup system so they are now in svn for community as well. Joshua D. Drake
On Mon, Feb 25, 2008 at 2:19 AM, Joshua D. Drake <jd@commandprompt.com> wrote: > Alvaro Herrera wrote: > > Alvaro Herrera wrote: > >> Magnus Hagander wrote: > > > >>> Aside from that, I assume you have verified that the search system still > >>> pulls things in properly? That's where I'd see a potential issue, but I > >>> don't have time to look into it right now (need to fix broken slides for > >>> tomorrows FOSDEM talks) > >> Good question -- I didn't. > > > > Ok, it seems to work. I searched for "archives.postgresql.org change" > > and this thread shows up. > > > > A note on the svn, the files are now in the autobackup system so they > are now in svn for community as well. Config files you mean? There shouldn't be any content type stuff in auto-backup, as that should work the other way round as people test changes in sandboxes and then commit them to SVN, from where the production site can check them out (something we'll need to sort out for archives). -- Dave Page EnterpriseDB UK: http://www.enterprisedb.com The Oracle-compatible database company
On Sun, Feb 24, 2008 at 4:39 PM, Joshua D. Drake <jd@commandprompt.com> wrote: > Magnus Hagander wrote: > > > As long as these things are covered, that's fine. But it'd be good if > > someone could verify it who actually knew about it. I think that means > > you, Marc :-) But if that's already done, then drop the point - it just > > wasn't clear from the original message. > > And Marc has been in the loop the whole time (just fyi). Then he also deserves a proverbial slap for not sharing with the rest of us. We are managing services used by thousands of people here - we need to coordinate all major changes to ensure everyone has a chance to think of any potential problems, and so we know whats happening. Consider; Stefan and I are chatting and notice that Nagios has started alerting on a service. We then spend an hour or two trying to diagnose the problem between us, not realising that the old box which we're looking at is fine, it's just that you've moved the service to a different server and something is subtly different, causing Nagios to barf. Communication JD - that's all we need!! -- Dave Page EnterpriseDB UK: http://www.enterprisedb.com The Oracle-compatible database company