"housekeeper" performance

[vrtareg]

Hi

I installed last housekeeper patch from Alexei and now my graphs looks like attached picture.

http://www.zabbix.com/forum/showthread.php?t=1811&page=1&pp=10&highlight=housekeeper

http://www.zabbix.com/forum/attachment.php?attachmentid=415

I found in log files that housekeeper runs for about 12 minutes and in that period zabbix server can't write data into database.

Server version: 1.1.1
Values stored 56723790
Trends stored 1644655
Number of hosts (monitored/not monitored/templates/deleted) 59(47/0/12/0)
Number of items (monitored/disabled/not supported)[trapper] 2341(1806/402/133)[0]
Number of triggers (enabled/disabled)[true/unknown/false] 300(296/4)[19/1/276]
Number of alarms 11227
Number of alerts 607

This is the log fragment for last execution.

026895:20060807:173204 Executing housekeeper
026905:20060807:174406 Timeout while receiving data from [srv62]
026905:20060807:174406 Getting value of [vfs.dev.write[hda,operations]] from host [srv62] failed
026905:20060807:174407 The value is not stored in database.
026895:20060807:174445 Deleted 488517 records from history and trends
026895:20060807:174445 Next housekeeper run is after 1h

Does anyone found the same problem?

[mike_k]

I can confirm such behaivour in 1.1.4
Using MyISAM tables, mysql 4.1.16

Any suggestions on how to reduce such timeouts?
Would it take much more time if it'll run once a day instead of once an hour? Or is it possible at least to set the exact time?

[vrtareg]

Hi

I solve this problem.
I move zabbix to new more powerfull server and change all zabbix tables to InnoDB.
Now i dont have such timeouts.

Areg

[mike_k]

Hi

I solve this problem.
I move zabbix to new more powerfull server and change all zabbix tables to InnoDB.
Now i dont have such timeouts.

Areg

Server is not that slow: load average: 0.52, 0.52, 0.45
Are there some reasons MyISAM engine results in such timeouts?

[vrtareg]

Yes

MyISAM uses another mechanism and all tables is going to be locked during housekeeper checks.
You should use InnoDB to avoid table locking.

Areg

[sbadger]

I have tried using the InnoDB, but the files grew to fast and I needed to bring the system down weekly to clean up the files. How are you dealing with the large files and maintining them?

[mike_k]

I have tried using the InnoDB, but the files grew to fast and I needed to bring the system down weekly to clean up the files. How are you dealing with the large files and maintining them?

large binary log?
you can run 'reset master;' periodically