Re: Huge increase of database partition - Mailing list pgsql-admin

From Thomas SIMON
Subject Re: Huge increase of database partition
Date
Msg-id 526FCE7F.5010907@neteven.com
Whole thread Raw
In response to Re: Huge increase of database partition  ("Plugge, Joe R." <JRPlugge@west.com>)
List pgsql-admin
You're right, indeed.
I saw same phenomenon today, and i have seen that the pgsql_tmp directory was full of temporary files linked to a particular query.
After i killed the query , postgres cleaned directory.
So, if i understand right, for avoid that the temp files crash data partition (and my db), i should create a new tablespace, and change param "temp_tablespaces" in postgresql configuration, right ?


Thomas
Le 24/10/2013 16:26, Plugge, Joe R. a écrit :

Yes, did you perhaps have a long running query that was doing lots of sorts, if so it creates as many 1 GB temp files as it takes to complete the sort, and they go into your data partition by default.  Once the query completes, postgres cleans them up.

 

From: pgsql-admin-owner@postgresql.org [mailto:pgsql-admin-owner@postgresql.org] On Behalf Of Thomas SIMON
Sent: Thursday, October 24, 2013 9:21 AM
To: pgsql-admin@postgresql.org
Subject: [ADMIN] Huge increase of database partition

 

Hi the list,

Yesterday, i have had a big problem of disk space on my server, running postgresql 9.0.13; with streaming replication.

Between 6pm and 7-30pm, i went from 45% used space to 99% (100% ?) used space on my /data partition (who contains my postgresql/9.0/main/base ; but does not contains the WAL; they are deported on an other partition). - seen only after with munin -

Juste after that - still according to munin- , the partition suddently decreases to its normal value (around 45%)
This peak represents around 165 Go of data ! ; beacause my partition size is 315 Go.

I can't explain why, it's the first time i have something like this. Usually, the size of the partition is very stable ; and grows very slowly.
Still with munin, i see that in this period, the blocks read per second increases from around 20k to 80k,  and i see that the disk utilisation is around 100%, which confirms that postgres wrote a lot of data.

I don't understand how this can happen, have you ever seen things like this ?

Thanks in advance,
Thomas

 


pgsql-admin by date:

Previous
From: Tom Lane
Date:
Subject: Re: displaying UTC time in local time
Next
From: JotaComm
Date:
Subject: Re: pg_dump being killed by oom killer