Re: temporary data after diskspace error

On 1/27/20 10:05 AM, Willy-Bas Loos wrote:
> Hi,
>
> We have a server with postgresql 9.4.12 on ubuntu.
> There has been a sudden rise in the amount of disk space used by
> postgresql, causing a diskspace error:
>
> 2020-01-22 17:24:37 CET db: ip: us: PANIC:  could not write to file
> "pg_xlog/xlogtemp.23346": No space left on device
> 2020-01-22 17:24:37 CET db: ip: us: LOG:  WAL writer process (PID 23346)
> was terminated by signal 6: Aborted
> 2020-01-22 17:24:37 CET db: ip: us: LOG:  terminating any other active
> server processes
>
> The disk was at roughly 75% before and something or someone added  150 GB
> to the database, bringing the disk space usage to 100%.
> The query that got the initial error was creating a rather large table,
> but it is not confirmed that this is the only source of the large-ish data
> amount. But it is possible.
>
> Now i can see in pg_stat_database and postgresql/9.4/main/base/pgsql_tmp
> that there is 90GB of temporary files in the database.
>
> Could the amount of temp files be caused by the unfinished query? I'm not
> sure how strong Signal 6 is exactly.
>
> And also: How can i make postgres clean up the files?
> Can it be done without restarting the cluster?
> Will restarting it help?

Restarting postgresql clears out pgsql_tmp.  "pg_ctl restart -D ..." should
be all you need.

--
Angular momentum makes the world go 'round.