Hi,
I fell into the following problem (unfortunately, the database contents
has sensitive customer information, so can publish very little of that).
Currently postgress process takes close to 100% CPU time.
I've restarted the process a moment ago, and it was calm for a brief
minute.
It started to consume the CPU once I've issued:
ALTER GROUP majster DROP USER rafal;
... which took close to 5minutes to complete. Then I've issued:
VACUUM ANALYZE ;
... which I had to cancel when it didnt' completed within another
10minutes. CPU utilisation remains at 100% even after the VACCUM was
canceled.
pg_dumpall was successfull during the large-CPU-usage time.
...another postgress-process restart, and another VACCUM ANALYSE, this
one completes in 3-5 secs. Now "ALTER GROUP ..." goes OK. And everything
seams working.
the database is currently almost empty: c.a. 100k records within c.a. 20
tables with another 30 or so views. There are c.a. 6k userids (roles)
created in the cluster.
Any ideas where should I look (system tables? process stats??) to
dieagnose the problem? .... just to be prepared when it appears again?
This is posgresql 8.1.4 on linux-debian.
--
-R