I have Zabbix server installed on Ubuntu 12.04.4 LTS (x86_64). This server has replaced old one which had Zabbix 1.4, the database was successfully converted to 2.2. For now, it has 176 hosts monitored, 1242 items, 1165 triggers. The server has 2 GB RAM, the swap is not used, load average is always below 0.5. "df -h" and "df -i" show that I have a lot of free space.
I had to tune zabbix to reduce queue when events are coming en-masse:
StartPollers=64
StartPingers=8
This made zabbix[queue] reduce from 800 to 30-40. This was done 4 months ago.
It worked fine for almost a year and then it stopped sending notifications. Everything stops at once: e-mail, jabber and sms. Somehow they get stuck in the database. If I restart zabbix_server or the whole server, all the notifications are sent to users, then it works for around a day or two and gets stuck again.
I've tried to manually alter database, deleting everything from escalations and alerts, but the best I could do is to eliminate the queue of outdated notifications before the server restarts.
The log file shows nothing suspicious. Just a spam of "cannot send list of active checks ... host not found". This is an issue with agents misconfiguration. (I still don't have free time to go through 50+ hosts and fix agents).
What should I do with notifications? I can't even receive simple ICMP ping messages like "Host %s is unreachable".
I had to tune zabbix to reduce queue when events are coming en-masse:
StartPollers=64
StartPingers=8
This made zabbix[queue] reduce from 800 to 30-40. This was done 4 months ago.
It worked fine for almost a year and then it stopped sending notifications. Everything stops at once: e-mail, jabber and sms. Somehow they get stuck in the database. If I restart zabbix_server or the whole server, all the notifications are sent to users, then it works for around a day or two and gets stuck again.
I've tried to manually alter database, deleting everything from escalations and alerts, but the best I could do is to eliminate the queue of outdated notifications before the server restarts.
The log file shows nothing suspicious. Just a spam of "cannot send list of active checks ... host not found". This is an issue with agents misconfiguration. (I still don't have free time to go through 50+ hosts and fix agents).
What should I do with notifications? I can't even receive simple ICMP ping messages like "Host %s is unreachable".
Comment