Обсуждение: Vacuum hangs
PG: 7.4.5
RH 7.3
Vac_Mem=12mg
Doing a Vacuum Analyze on a table and it's been hanging for at least 2 hours. I could see no other users hitting the table at the time. This is the second time this has happened on this particular table and I can't seem to pin point the issue. If I kill the PID and restart the vacuum analyze it works fine. Any suggestion where I start searching? Here is an output from pg_stat_activity and the
mdc_oz=# select * from pg_stat_activity:
datid | 17145
datname | mdc_oz
procpid | 16560
usesysid | 2
usename | phatcher
current_query | VACUUM ANALYZE cdm.cdm_reg_customer;
query_start | 2004-11-09 10:02:32.523146-08
mdc_oz=# select * from pg_locks where pid=16560;
-[ RECORD 1 ]-------------------------
relation | 7059380
database | 17145
transaction |
pid | 16560
mode | ShareUpdateExclusiveLock
granted | t
-[ RECORD 2 ]-------------------------
relation |
database |
transaction | 28863986
pid | 16560
mode | ExclusiveLock
granted | t
-[ RECORD 3 ]-------------------------
relation | 7059380
database | 17145
transaction |
pid | 16560
mode | ShareUpdateExclusiveLock
granted | t
TIA
Patrick Hatcher
Patrick Hatcher <PHatcher@macys.com> writes:
> Doing a Vacuum Analyze on a table and it's been hanging for at least 2
> hours. I could see no other users hitting the table at the time. This is
> the second time this has happened on this particular table and I can't
> seem to pin point the issue. If I kill the PID and restart the vacuum
> analyze it works fine. Any suggestion where I start searching?
Next time, please attach to the vacuuming backend with gdb and get a
stack trace. The pg_locks output doesn't tell us anything except that
the problem isn't a lock :-(
regards, tom lane
Will do thanks. I ended up dropping and recreating the index and that
seems to have fixed the problem.
Patrick Hatcher
Tom Lane
<tgl@sss.pgh.pa.u
s> To
Patrick Hatcher
11/09/2004 12:06 <PHatcher@macys.com>
PM cc
pgsql-general@postgresql.org
Subject
Re: [GENERAL] Vacuum hangs
Patrick Hatcher <PHatcher@macys.com> writes:
> Doing a Vacuum Analyze on a table and it's been hanging for at least 2
> hours. I could see no other users hitting the table at the time. This
is
> the second time this has happened on this particular table and I can't
> seem to pin point the issue. If I kill the PID and restart the vacuum
> analyze it works fine. Any suggestion where I start searching?
Next time, please attach to the vacuuming backend with gdb and get a
stack trace. The pg_locks output doesn't tell us anything except that
the problem isn't a lock :-(
regards, tom lane