Sorry folks, I’m still home nursing a nasty chest cold and my only tool today is an iPad.
I have failed to get the postgresql.conf into the copy buffer so that, along with the results of pg_settings, will have to wait for another day.
Today there are “only” 135 Mar 16 WAL files. I haven’t sorted out which have been cleaned up but can do so if that’s thought to be helpful.
There is still 2.2G in the pg_wal directory but that disc has ~360G left. (I believe the burst of WAL files was the result of a novice using LIMIT with a to-Json function and the target table has >100M rows.)
Given that current WALs come and go regularly, I think the CHECKPOINT is running frequently enough (for normal loads at least).
> On Mar 30, 2019, at 10:54 AM, Gmail <robjsargent@xxxxxxxxx> wrote:
>
>
>>>> On Mar 29, 2019, at 6:58 AM, Michael Paquier <michael@xxxxxxxxxxx> wrote:
>>>
>>> On Thu, Mar 28, 2019 at 09:53:16AM -0600, Rob Sargent wrote:
>>> This is pg10 so it's pg_wal. ls -ltr
>>>
>>>
>>> -rw-------. 1 postgres postgres 16777216 Mar 16 16:33
>>> 0000000100000CEA000000B1
>>> -rw-------. 1 postgres postgres 16777216 Mar 16 16:33
>>> 0000000100000CEA000000B2
>>>
>>> ... 217 more on through to ...
>>>
>>> -rw-------. 1 postgres postgres 16777216 Mar 16 17:01
>>> 0000000100000CEA000000E8
>>> -rw-------. 1 postgres postgres 16777216 Mar 16 17:01
>>> 0000000100000CEA000000E9
>>> -rw-------. 1 postgres postgres 16777216 Mar 28 09:46
>>> 0000000100000CEA0000000E
> I’m now down to 208 Mar 16 WAL files so they are being processed (at least deleted). I’ve taken a snapshot of the pg_wal dir such that I can see which files get processed. It’s none of the files I’ve listed previously
Two more have been cleaned up. 001C and 001D generated at 16:38 Mar 16
Please share your complete postgresql.conf file and the results from this query:
SELECT * FROM pg_settings;
has someone in the past configured wal archiving?
You've ran out of disk space as this log message you shared states:
No space left on device
what's the output of df -h
--
BTW , how spread apart are checkpoints happening? do you have stats on that? maybe they're too spread apart and that's why WAL files cannot be recycled rapidly enough?
--
