Re: Better handling of archive_command problems - Mailing list pgsql-hackers

From Daniel Farina
Subject Re: Better handling of archive_command problems
Date
Msg-id CAAZKuFbJ+A+xahG-VD9DuNnFXoHZkLeyOvtTt0bZsOnm6V4Vmg@mail.gmail.com
Whole thread Raw
In response to Re: Better handling of archive_command problems  (Robert Haas <robertmhaas@gmail.com>)
List pgsql-hackers
On Thu, May 16, 2013 at 9:13 PM, Robert Haas <robertmhaas@gmail.com> wrote:
> On Thu, May 16, 2013 at 10:06 PM, Daniel Farina <daniel@heroku.com> wrote:
>> Do you have a sketch about mechanism to not encounter that problem?
>
> I didn't until just now, but see my email to Peter.  That idea might
> be all wet, but off-hand it seems like it might work...
>
>> However little it may matter, I would like to disagree with your
>> opinion on this one: the current situation as I imagine encountered by
>> *all* users of archiving is really unpleasant, 'my' shop or no.  It
>> would probably not be inaccurate to say that 99.9999% of archiving
>> users have to live with a hazy control over the amount of data loss,
>> only bounded by how long it takes for the system to full up the WAL
>> file system and then for PostgreSQL to PANIC and crash (hence, no more
>> writes are processed, and no more data can be lost).
>
> I'm really not trying to pick a fight here.  What I am saying is that
> it is the problem is not so bad that we should accept the first design
> proposed, despite the obvious problems with it, without trying to find
> a better one.  The first post to -hackers on this was 4 days ago.
> Feature freeze for the first release that could conceivably contain
> this feature will most likely occur about 8 months from now.  We can
> take a few more days or weeks to explore alternatives, and I believe
> it will be possible to find good ones.  If I were trying to kill this
> proposal, I could find better ways to do it than clearly articulating
> my concerns at the outset.

I know you are not trying to kill it or anything, and I didn't mean to
pick a fight either: my point is, without rancor, that I think that
as-is the state of affairs fails the sniff test dimensions like
availability and exposure to data loss, much worse than some of the
negative effects of hobbling cancellations.

And I completely agree with you that given the timing of thinking
about this issue means one can marinate it for a while.

And, I hope it goes without saying, I am grateful for your noodling on
the matter, as always.



pgsql-hackers by date:

Previous
From: Robert Haas
Date:
Subject: Re: Better handling of archive_command problems
Next
From: Robert Haas
Date:
Subject: Re: Extent Locks