Zabbix 7.0.26 - 7.0.29 housekeeper consumes all available RAM
Hello,
I have a memory problem with the housekeeper in Zabbix 7.0.26 - 7.0.29 using MariaDB.
Housekeeping was disabled for some time, so old history data accumulated. After starting it manually, the housekeeper continuously consumes more RAM while processing old history and trends:
1281 11303284 11896316 /usr/sbin/zabbix_server: housekeeper [removing old history and trends]
1347 7080 599644 /usr/sbin/zabbix_server: trigger housekeeper [deleted 0 problems records in 0.013036 sec, idle for 60 second(s)]
The VM originally had 8 GB RAM and became unresponsive. I increased it to 16 GB, but the housekeeper consumed that memory as well. More RAM only delays the problem. After restarting zabbix-server, the memory is released and the server remains stable until housekeeping is started again.
This seems related to the initial history queries that retrieve itemid and MIN(clock) from all history and trend tables.
Is this a known issue in Zabbix 7.0? Is there a way to limit the housekeeper’s memory usage or safely process the accumulated history in smaller batches?
Environment:
Best regards
I have a memory problem with the housekeeper in Zabbix 7.0.26 - 7.0.29 using MariaDB.
Housekeeping was disabled for some time, so old history data accumulated. After starting it manually, the housekeeper continuously consumes more RAM while processing old history and trends:
1281 11303284 11896316 /usr/sbin/zabbix_server: housekeeper [removing old history and trends]
1347 7080 599644 /usr/sbin/zabbix_server: trigger housekeeper [deleted 0 problems records in 0.013036 sec, idle for 60 second(s)]
The VM originally had 8 GB RAM and became unresponsive. I increased it to 16 GB, but the housekeeper consumed that memory as well. More RAM only delays the problem. After restarting zabbix-server, the memory is released and the server remains stable until housekeeping is started again.
This seems related to the initial history queries that retrieve itemid and MIN(clock) from all history and trend tables.
Is this a known issue in Zabbix 7.0? Is there a way to limit the housekeeper’s memory usage or safely process the accumulated history in smaller batches?
Environment:
- Zabbix 7.0.29
- Debian 13
- MariaDB
- 16 GB RAM
Best regards