Re: Would like to know how analyze works technically
| От | TonyS |
|---|---|
| Тема | Re: Would like to know how analyze works technically |
| Дата | |
| Msg-id | c1a39a9a7d771cd42f65dee5a7c072ab.squirrel@www.exquisiteimages.com обсуждение исходный текст |
| Ответ на | Re: Would like to know how analyze works technically (Kevin Grittner <kgrittn@ymail.com>) |
| Ответы |
Re: Would like to know how analyze works technically
(Tom Lane <tgl@sss.pgh.pa.us>)
|
| Список | pgsql-general |
On Wed, April 1, 2015 9:35 am, Kevin Grittner-5 [via PostgreSQL] wrote:
>
>
> TonyS <[hidden email]> wrote:
>
>
>> The postgresql log has these entries at the crash point:
>> 2015-04-01 06:24:37 EDT LOG: server process (PID 1384) was terminated
>> by signal 9: Killed 2015-04-01 06:24:38 EDT DETAIL: Failed process was
>> running: analyze verbose;
>>
>
> That was almost certainly the action of the OS's Out Of Memory
> Killer process.
>
>
>> Is there anything else that would be helpful?
>>
>
> Unfortunately, leaving the OOM killer enabled causes the best
> evidence to be destroyed. If you disable the OOM killer and run this
> again, when memory is exhausted the database process attempting to
> allocate memory will dump a map of where its memory was allocated. That
> should give us something to work with regarding the cause. Try:
>
> vm.overcommit_memory = 2 vm.overcommit_ratio = 80
>
The analyze function has crashed again while the overcommit entries were
as above. The last bit of the PostgreSQL log shows:
MdSmgr: 41934848 total in 14 blocks; 639936 free (0 chunks); 41294912 used
ident parser context: 0 total in 0 blocks; 0 free (0 chunks); 0 used
hba parser context: 7168 total in 3 blocks; 2288 free (1 chunks); 4880 used
LOCALLOCK hash: 8192 total in 1 blocks; 1680 free (0 chunks); 6512 used
Timezones: 83472 total in 2 blocks; 3744 free (0 chunks); 79728 used
ErrorContext: 8192 total in 1 blocks; 8160 free (6 chunks); 32 used
2015-04-01 14:23:27 EDT ERROR: out of memory
2015-04-01 14:23:27 EDT DETAIL: Failed on request of size 80.
2015-04-01 14:23:27 EDT STATEMENT: analyze verbose;
From syslog leading up to the crash
Apr 1 06:24:37 postgrestest kernel: [26692.691936] [ 1434] 1000 1434
26408 70 53 197 0 sshd
Apr 1 06:24:37 postgrestest kernel: [26692.691939] [ 1435] 1000 1435
5593 0 16 452 0 bash
Apr 1 06:24:37 postgrestest kernel: [26692.691943] [ 1449] 1000 1449
6204 81 17 31 0 top
Apr 1 06:24:37 postgrestest kernel: [26692.691947] Out of memory: Kill
process 1384 (postgres) score 790 or sacrifice child
Apr 1 06:24:37 postgrestest kernel: [26692.693667] Killed process 1384
(postgres) total-vm:10044764kB, anon-rss:6963780kB, file-rss:202916kB
Apr 1 06:25:02 postgrestest CRON[1755]: (root) CMD (test -x
/usr/sbin/anacron || ( cd / && run-parts --report /etc/cron.daily ))
View this message in context: Re: Would like to know how analyze works technically
Sent from the PostgreSQL - general mailing list archive at Nabble.com.
>
>
> TonyS <[hidden email]> wrote:
>
>
>> The postgresql log has these entries at the crash point:
>> 2015-04-01 06:24:37 EDT LOG: server process (PID 1384) was terminated
>> by signal 9: Killed 2015-04-01 06:24:38 EDT DETAIL: Failed process was
>> running: analyze verbose;
>>
>
> That was almost certainly the action of the OS's Out Of Memory
> Killer process.
>
>
>> Is there anything else that would be helpful?
>>
>
> Unfortunately, leaving the OOM killer enabled causes the best
> evidence to be destroyed. If you disable the OOM killer and run this
> again, when memory is exhausted the database process attempting to
> allocate memory will dump a map of where its memory was allocated. That
> should give us something to work with regarding the cause. Try:
>
> vm.overcommit_memory = 2 vm.overcommit_ratio = 80
>
The analyze function has crashed again while the overcommit entries were
as above. The last bit of the PostgreSQL log shows:
MdSmgr: 41934848 total in 14 blocks; 639936 free (0 chunks); 41294912 used
ident parser context: 0 total in 0 blocks; 0 free (0 chunks); 0 used
hba parser context: 7168 total in 3 blocks; 2288 free (1 chunks); 4880 used
LOCALLOCK hash: 8192 total in 1 blocks; 1680 free (0 chunks); 6512 used
Timezones: 83472 total in 2 blocks; 3744 free (0 chunks); 79728 used
ErrorContext: 8192 total in 1 blocks; 8160 free (6 chunks); 32 used
2015-04-01 14:23:27 EDT ERROR: out of memory
2015-04-01 14:23:27 EDT DETAIL: Failed on request of size 80.
2015-04-01 14:23:27 EDT STATEMENT: analyze verbose;
From syslog leading up to the crash
Apr 1 06:24:37 postgrestest kernel: [26692.691936] [ 1434] 1000 1434
26408 70 53 197 0 sshd
Apr 1 06:24:37 postgrestest kernel: [26692.691939] [ 1435] 1000 1435
5593 0 16 452 0 bash
Apr 1 06:24:37 postgrestest kernel: [26692.691943] [ 1449] 1000 1449
6204 81 17 31 0 top
Apr 1 06:24:37 postgrestest kernel: [26692.691947] Out of memory: Kill
process 1384 (postgres) score 790 or sacrifice child
Apr 1 06:24:37 postgrestest kernel: [26692.693667] Killed process 1384
(postgres) total-vm:10044764kB, anon-rss:6963780kB, file-rss:202916kB
Apr 1 06:25:02 postgrestest CRON[1755]: (root) CMD (test -x
/usr/sbin/anacron || ( cd / && run-parts --report /etc/cron.daily ))
View this message in context: Re: Would like to know how analyze works technically
Sent from the PostgreSQL - general mailing list archive at Nabble.com.
В списке pgsql-general по дате отправления: