You have a very large number of jobs all trying to do inserts into the 
database at once.  What you are probably hitting is the limit of the 
number of transactions/second that your database server can sustain - 
most likely you are I/O bound at the disk level.

All of those 'inserting attribute' jobs are asking for many, many small 
I/O operations (IOPs) - since they are almost certainly not being 
written sequentially this will be causing a lot of seeks on the drive.  
Basically the way you are doing things is probably causing the head on 
the disk to chatter around like crazy.

You will probably find your jobs complete faster if you run fewer at 
once, as right now you are spending most of your time waiting for the 
heads to move on the disks rather than having them actually write anything.

If you run 'iostat -x 2' you will be able to see the %UTL for your 
physical disks (assuming you are on Linux) - you'll probably see your 
disks are in the high 90% if not 100% when all those 'Dir inserting 
Attributes' jobs are running.  As I mentioned above, what you are seeing 
is almost certainly the IOP limits of your hardware rather than the 
throughput limits.

Bryn

On 14-10-20 02:14 AM, Anton Gorlov wrote:
> The problem is observed recently time.
>
> After database backup "Dir inserting Attributes" job runs more than 5 hours.
> Previously, it takes less time. Size of database changed slightly, current 
> size
> - 85Gb.
> the loads of io system: iotop read 5 M/s, write several Kb
> LA 1.4-2.2
> atop retruns postgres process with bacula INSERT ... argument
>
> hdparm -tT /dev/mapper/main-pgsql
>
> /dev/mapper/main-pgsql:
>    Timing cached reads:   14626 MB in  2.00 seconds = 7321.24 MB/sec
>    Timing buffered disk reads: 918 MB in  3.03 seconds = 302.72 MB/sec
>
>
>
> Request list contains:
> BEGIN; LOCK TABLE Path IN SHARE ROW EXCLUSIVE MODE
>
>
> ===
> Running Jobs:
> Console connected at 20-Ок-2014 11:52
>    JobId Level   Name                       Status
> ======================================================================
>     8116 Full    fluorine.2014-10-19_00.25.00_40 Dir inserting Attributes
>     8117 Full    hydrogen.2014-10-19_00.25.00_41 Dir inserting Attributes
>     8118 Full    neon.2014-10-19_00.32.01_42 Dir inserting Attributes
>     8122 Full    aluminium_DB0.2014-10-19_02.30.00_46 Dir inserting Attributes
>     8124 Full    aluminium_DB5.2014-10-19_02.35.00_48 Dir inserting Attributes
>     8127 Full    aluminium_DB2.2014-10-19_02.50.01_51 Dir inserting Attributes
>     8129 Differe  hydrogen.2014-10-20_00.50.00_57 is waiting on max Client 
> jobs
>     8130 Differe  sulfur_bcp.2014-10-20_01.10.00_01 Dir inserting Attributes
>     8131 Differe  boron_bcp.2014-10-20_02.00.00_03 Dir inserting Attributes
>     8132 Differe  neon.2014-10-20_02.05.00_04 Dir inserting Attributes
>     8133 Full    aluminium_PG5.2014-10-20_02.10.00_05 Dir inserting Attributes
>     8134 Differe  fluorine.2014-10-20_02.20.00_06 Dir inserting Attributes
>     8135 Full    aluminium_DB0.2014-10-20_02.30.00_07 Dir inserting Attributes
>     8136 Full    aluminium_DB4.2014-10-20_02.34.00_08 Dir inserting Attributes
>     8137 Full    aluminium_DB5.2014-10-20_02.35.00_09 Dir inserting Attributes
>     8138 Full    phosphorus_bcp.2014-10-20_02.35.00_10 Dir inserting 
> Attributes
>     8139 Full    aluminium_PG0.2014-10-20_02.35.00_11 Dir inserting Attributes
>     8140 Full    aluminium_PG2.2014-10-20_02.35.00_12 Dir inserting Attributes
>     8141 Full    aluminium_DB2.2014-10-20_02.50.00_13 Dir inserting Attributes
>     8142 Full    aluminium_BILL.2014-10-20_04.35.00_14 Dir inserting 
> Attributes
> ====
>
>
> postgresql.conf:
> max_connections = 100
> shared_buffers = 128MB
> temp_buffers = 512MB
> work_mem = 64MB
> maintenance_work_mem = 32MB
> fsync = off
> checkpoint_segments = 90
>
>
>
> ------------------------------------------------------------------------------
> Comprehensive Server Monitoring with Site24x7.
> Monitor 10 servers for $9/Month.
> Get alerted through email, SMS, voice calls or mobile push notifications.
> Take corrective actions from your mobile device.
> http://p.sf.net/sfu/Zoho
> _______________________________________________
> Bacula-users mailing list
> Bacula-users AT lists.sourceforge DOT net
> https://lists.sourceforge.net/lists/listinfo/bacula-users


------------------------------------------------------------------------------
Comprehensive Server Monitoring with Site24x7.
Monitor 10 servers for $9/Month.
Get alerted through email, SMS, voice calls or mobile push notifications.
Take corrective actions from your mobile device.
http://p.sf.net/sfu/Zoho
_______________________________________________
Bacula-users mailing list
Bacula-users AT lists.sourceforge DOT net
https://lists.sourceforge.net/lists/listinfo/bacula-users