Zabbix 6.0 SNMP still timeouts

[rvandenzel]

Hi All,

We've recently installed Zabbix 6.0 and added out storage hosts (Dell/EMC Compellent), but if we try to monitor them, SNMP says 'Timeout while connecting to '<IP-address>'.

If I do an snmpwalk from the Zabbix server it produces output:

snmpwalk -v3 -l authPriv -u <zabbix user on storage> -a SHA -A ******* -x AES -X ******* <ip-address of compellent> .1.3.6.1.4.1.16139.2.27.1.4
SNMPv2-SMI::enterprises.16139.2.27.1.4.1 = STRING: "*********"

Also nc works fine:

nc -z -u -v <ip-address of compellent> 161
Connection to <ip-address of compellent> 161 port [udp/snmp] succeeded!

Our Zabbix 6.0 server is running on RHEL 8.5, no SElinux of Firwall enabled on it (yet).

Please help or advice in where I'm thinking wrong!

Regards,

Richard

[rvandenzel]

Anyone for help?

[cyber]

Just throwing some options in the air... Do yo have multiple addresses attached to your zabbix server? Maybe server uses different IP for outgoing connections that you from command line?
Your host is monitored directly from server and no proxy?

[rvandenzel]

Nope, Zabbix-servers have only 1 address assigned to them. We are not using Zabbix proxy, so yes, directly monitored from server.

[nklocalit]

Did you ever manage to resolve this ? I have similar problem. After server restart it works fine sometimes, but after a while it keeps timeout-ing ...

[tim.mooney]

What is your "Timeout" setting in /etc/zabbix/zabbix_server.conf or related config file(s) from /etc/zabbix/zabbix_server.d/*.conf ?

Zabbix (server) uses a bunch of different processes to perform different tasks. You can see many of the worker "types" in the process list when you use "ps" on your Zabbix server. You can increase the log level of individual worker types using the "runtime control" options from zabbix_server(8) man page. What I would do in your situation is temporarily increase the log level for just the "poller" worker type, and then dig through the resulting log entries to see if the increased verbosity of the poller entries gives you any idea what the problem is. Assuming you use "sudo" for administrative access, the command would be something like

sudo -u zabbix zabbix_server -R log_level_increase=poller

You might need to run that command twice, to increase the log level by 2 levels.

Be sure you remember to undo that after you've gathered your extra log entries.

That's what I would try, if I were in your situation.

[nklocalit]

What is your "Timeout" setting in /etc/zabbix/zabbix_server.conf or related config file(s) from /etc/zabbix/zabbix_server.d/*.conf ?

I've played with various "Timeout" settings. I've tried 5, 10 and 30s but it doesn't make any difference.
Current setup for Zabbix Server is very "relaxed" and shows to use very little resources.
I should mention that it runs on docker and those problematic (timeout-ed) SNMPs are coming from discovered hosts (those are wireless Cisco APs discovered out of the Cisco WLC controller).
When I run SNMP walk everything works fine, but in Zabbix after a while it just errors keep timeouting. Other Cisco devices (switches, routers etc) on network are working fine...

Thanks for suggestions.

Kind regards.

[scottmonet]

I'm in the same boat - snmpwalk works fine from the server (and to other devices on the same network segment), but the polling throws the error message 'Timeout while connecting to '<IP-address>'.
I'll keep looking. It's happened before when adding new devices, but I update the connection settings and the error has just gone away in the past.
Good luck! And I'll post any solutions that I come up with.

[ABALONE]

I have the same problem the snmp timeout occasionally for over hundred IoT!
1. I use the env variable increase the poller and etc, it does help but after a few days i realize the snmp timeout getting worse!
2. To further verify, I move the zabbix away from container to a native env in a RPI. Damn it works like a charm in a native environment!
Frankly speaking, i suspect the docker has a network resource constraint on zabbix server. I use the wireshark to inspect all the packet on snmp i realize the snmp request is not send by the zabbix server.
Just move away from container that is my suggestion.