Search Postgresql Archives

Re: very slow queries and ineffective vacuum

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hello Lukasz,

On Tue, Jun 30, 2015 at 8:57 AM, Lukasz Wrobel <lukasz.wrobel@xxxxxxxxxxxxxxxxxxxxx> wrote:

There doesn't seem to be any issues with disk space, memory or CPU, as neither of those is even 50% used (as per df and top).

Are you using the default PostgreSQL configuration settings, or have you custom tuned them? The default settings are targeted for wide compatibility and are not optimized for performance. If PostgreSQL is performing badly and using a small amount of system resources it is likely some tuning is needed. See docs: http://www.postgresql.org/docs/current/static/runtime-config.html 


On Tue, Jun 30, 2015 at 8:57 AM, Lukasz Wrobel <lukasz.wrobel@xxxxxxxxxxxxxxxxxxxxx> wrote:

For whatever reason there is also no data in pg_stat* tables.

You can also turn on tracking (for statistics views) by enabling statistics collection in the config http://www.postgresql.org/docs/current/static/runtime-config-statistics.html

Will J. Dunn

On Tue, Jun 30, 2015 at 8:57 AM, Lukasz Wrobel <lukasz.wrobel@xxxxxxxxxxxxxxxxxxxxx> wrote:
Hello.

I have multiple problems with my database, the biggest of which is how to find out what is actually wrong.

First of all I have a 9.3 postgres database that is running for about a month. Right now the queries on that database are running very slowly (select with a simple "where" on a non-indexed column on a table with about 5000 records takes 1,5s, a complicated hibernate select with 7 joins on tables of about 5000 records takes about 15s, insert or update on a table with 35000 records takes up to 20 mins).

The tables and indexes on those tables are bloated to the point where this query: https://wiki.postgresql.org/wiki/Show_database_bloat shows wasted bytes in hundreds of MB.

For whatever reason there is also no data in pg_stat* tables.

So due to the long query times, there are multiple errors in my application logs like "No free connection available" or "Could not synchronize database state with session", or "Failed to rollback transaction" and the application fails to start in the required time.

The only thing that helps fix the situation seems to be vacuum full of the entire database. Regular vacuum doesn't even lower the dead tuples count (which appear by the thousands during application launching). Reindex of all the indexes in the database didn't help as well. All autovacuum parameters are default.

There doesn't seem to be any issues with disk space, memory or CPU, as neither of those is even 50% used (as per df and top).

Is there any good tool that will monitor the queries and generate a report with useful information on what might be the problem? I tried pg_badger, but all I got were specific queries and their times, but the long query times are just one of the symptoms of what's wrong with the database, not the cause.

Perhaps I'm missing some indexes on the tables (creating them on the columns on which the where clause was used in the long queries seemed to halve their times). Also how can I monitor my transactions and if they are closed properly?

I will be grateful for any help and if you need more details I can provide them if possible.

Best regards.
Lukasz


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Index of Archives]     [Postgresql Jobs]     [Postgresql Admin]     [Postgresql Performance]     [Linux Clusters]     [PHP Home]     [PHP on Windows]     [Kernel Newbies]     [PHP Classes]     [PHP Books]     [PHP Databases]     [Postgresql & PHP]     [Yosemite]
  Powered by Linux