Thread: Age of the WAL?

Age of the WAL?

From
Erik Jones
Date:
What's the best way to determine the age of the current WAL?  Not the current segment, but the whole thing.  Put
anotherway:  is there a way to determine a timestamp for the oldest available transaction in the WAL? 

Re: Age of the WAL?

From
Tom Lane
Date:
Erik Jones <ejones@engineyard.com> writes:
> What's the best way to determine the age of the current WAL?  Not the current segment, but the whole thing.  Put
anotherway:  is there a way to determine a timestamp for the oldest available transaction in the WAL? 

Transaction commit and abort records carry timestamps, so you could
figure this out with something like pg_xlogdump.  I don't know of any
canned solution though.

            regards, tom lane


Re: Age of the WAL?

From
Erik Jones
Date:
On Mar 12, 2013, at 4:13 PM, Tom Lane wrote:

> Erik Jones <ejones@engineyard.com> writes:
>> What's the best way to determine the age of the current WAL?  Not the current segment, but the whole thing.  Put
anotherway:  is there a way to determine a timestamp for the oldest available transaction in the WAL? 
>
> Transaction commit and abort records carry timestamps, so you could
> figure this out with something like pg_xlogdump.  I don't know of any
> canned solution though.


Tom,

Thanks, and sorry for any discontinuity in the rather long time it's taken for me to get on this reply (had a
vacation).

Anyway, will pg_xlogdump work with any previous versions of Postgres or will it be only 9.3+?

For reference, the reason need to be able to do this is this:  Given a set of snapshots (each taken with running
pg_start_backupbefore and pg_stop_backup after) and a running server, I need to be able to determine whether or not any
givensnapshot will be usable for setting up a new standby.  I know I could grab the info returned from pg_start_backup
andstore that as snapshot meta-data but I'm hoping to keep from having to make changes the existing snapshotting code,
ifpossible. 

Re: Age of the WAL?

From
Tom Lane
Date:
Erik Jones <ejones@engineyard.com> writes:
> On Mar 12, 2013, at 4:13 PM, Tom Lane wrote:
>> Transaction commit and abort records carry timestamps, so you could
>> figure this out with something like pg_xlogdump.  I don't know of any
>> canned solution though.

> Anyway, will pg_xlogdump work with any previous versions of Postgres or will it be only 9.3+?

The version recently added to contrib is only meant to work with the
current server release, AFAIK.  However, it's derived from older
standalone programs that are out there somewhere --- did you look around
on pgfoundry?

            regards, tom lane


Re: Age of the WAL?

From
Peter Geoghegan
Date:
On 26 March 2013 22:21, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> The version recently added to contrib is only meant to work with the
> current server release, AFAIK.  However, it's derived from older
> standalone programs that are out there somewhere --- did you look around
> on pgfoundry?

Actually, I think the version on pgfoundry is unmainted.

I'd look here instead:

https://github.com/snaga/xlogdump/commits/master


--
Regards,
Peter Geoghegan