Hello,
Thanks for answering, sorry I get back to you so late but I was
testing configs.
On 19/04/2022 10:30, Mark Patruck wrote:
On 16.04.2022 20:44, Noth wrote:
Hi,
Hi,
I built the new 7.1 packages for Zabbix and PostgreSQL, and
upgraded my monitoring VM. To my horror the zabbix_server process
stops graphing after a couple of hours of uptime, with the
housekeeper and history syncer processes at over 80% cpu usage.
PostgreSQL shows INSERT processes stuck at 80% too, and restarting
zabbix_server hangs, leaving zombie processes. I was using 2 vcpus
and 2G of RAM. Feeling that the login.conf limits might be the
problem I uped them by quite a bit:
The zabbix_server process won't eat up your system normally...it's more
database related.
1) how big is your database? (important, as you don't have much memory
for a busy PostgreSQL database)
It's 5G (I've been running zabbix for years)
2) did you make changes to /var/postgresql/data/postgresql.conf?
(work_mem, shared_buffers, effective_cache_size, random_page_cost)
No I'm using the default
2) how many items do you monitor?
31
3) did you always follow the Zabbix upgrade guide (especially 5->6)?
I did, and did the special extra steps for postgresql
4) how do your Zabbix server graphs "data gathering", "internal",
"data handling" look like before the problem starts? Any high
spikes?
CPU goes to 100%, housekeeper process goes over 80% and top shows
postgresql processes for zabbix using near to 90% and they build up
(from two to about five-six) till zabbix_server fails with out of memory
(warning emails generated by zabbix don't go out because of this despite
regular system emails going out)
5) have you already tried with a new database instance, i.e. move
/var/postgresql/data to /var/postgresql/data_old and start with
Zabbix 6.0.3 from scratch to see, if the database issues start
immediately even when almost empty?
I did and it runs fine for 24 hours, using the initial ram+cpu config I had.
