Hello All, I'm trying to troubleshoot issues with the housekeeper not being able to keep up...
First off, a bit of background info:
We're monitoring 300 hosts, 40,000 items total across thoses hosts, monitored by 13 proxies, backed by PostGres
We've been running into issues with postgres running out of space and crashing.
The first issue we discovered was that we had misconfigured Zabbix to store 365 days of history. We shortened it to 15d
The second issue we discovered was that the growth of the history tables had never been sharp enough to trigger Postgres's auto-vacuum. we tweaked autovac to be more aggressive, and that now seems to be fixed
The 3rd issue we encountered was that the Zabbix Housekeeper was taking hours to complete, and only deleting a few thousand rows at a time. We found the Zabbix server conf file "MaxHousekeeperDelete" value, and set it to 0 (aka infinite)
The issue we now face is that housekeeper takes 4 hours to complete, and consistently only deletes 8 million records.
My question is, is there some additional undocumented limit preventing housekeeper from deleting more than 8 million rows per run?
First off, a bit of background info:
We're monitoring 300 hosts, 40,000 items total across thoses hosts, monitored by 13 proxies, backed by PostGres
We've been running into issues with postgres running out of space and crashing.
The first issue we discovered was that we had misconfigured Zabbix to store 365 days of history. We shortened it to 15d
The second issue we discovered was that the growth of the history tables had never been sharp enough to trigger Postgres's auto-vacuum. we tweaked autovac to be more aggressive, and that now seems to be fixed
The 3rd issue we encountered was that the Zabbix Housekeeper was taking hours to complete, and only deleting a few thousand rows at a time. We found the Zabbix server conf file "MaxHousekeeperDelete" value, and set it to 0 (aka infinite)
The issue we now face is that housekeeper takes 4 hours to complete, and consistently only deletes 8 million records.
My question is, is there some additional undocumented limit preventing housekeeper from deleting more than 8 million rows per run?