Centos 6/x86-64

Sounds not like a bug report.
You may find community support on IRC and forums and commercial support here

this might be very hard to trace down - i don't see anything that could be done right now.

next time this happens, you could try stracing poller processes and seeing whether any of the is stuck - and if so, on what.

Marc , not bug report?
I hit bug, and reported about it.
Still not? :-D

richlv, I'll definitely trace

"Problem" might appear more appropriate to me than "bug"

I'm sure this is bug . Period.

Sles, would be correct to provide from very beginning zabbix_server.conf's UnreachablePeriod and UnavailableDelay

Check this page https://www.zabbix.com/documentation/2.2/manual/appendix/items/unreachability you can find there some useful points.

Oleksiy Zagorskyi , they are default.
Anyway, there was no such "temporarily disabling Zabbix agent checks on host [New host]: host unavailable" message in log, only what I already provided.
There was no info in web interface that host in not monitored too.

Hello!

Just upgraded zabbix agent on several hosts, and on all of them got about 10 minites of no data:

14580:20131231:075602.521 Zabbix agent item "net.if.in[eth0,bytes]" on host "inetgw-nsk" failed: first network error, wait for 15 seconds
14589:20131231:080515.459 resuming Zabbix agent checks on host "inetgw-nsk": connection restored

14571:20131231:080655.216 Zabbix agent item "vfs.fs.size[/,used]" on host "ast-ngdu2.xnet.belkam.com" failed: first network error, wait for 15 seconds
14592:20131231:081515.430 resuming Zabbix agent checks on host "ast-ngdu2.xnet.belkam.com": connection restored

etc.

Why so long timeout?

Thank you!

Did you already:

• checked internal items
• checked health of your database
• inspected log files having DebugLevel set to 4
• investigated tcp traffic between involved components
• followed suggestions provided on Zabbix.org: Zabbix says "Agent unreachable", but the host is up. How do I debug that?

partially

1. yes
2. yes
3. not yet, will turn debug on , if problem occur again- I'll have no logs
4. no, because I still don't know how to reproduce problem
5. no, because this is definitely not agent problem

ok. about 4- just rebooted one of servers.
no zabbix agent related traffic (only snmp and icmp) beetwen zabbix server and server with agent for last 5 minutes.
zabbix shows host status Под наблюдением

will turn debug on later (quite busy right now) and try to reproduce.

now connection is restored:

23283:20140102:075108.042 Zabbix agent item "net.if.in[eth0,bytes]" on host "asterisk.p98.belkam.com" failed: first network error, wait for 15 seconds
23328:20140102:075231.300 item [asterisk.p98.belkam.com:calls] became not supported: SNMP error: (noSuchName) There is no such variable name in this MIB.
23328:20140102:075631.078 item [asterisk.p98.belkam.com:calls] became supported
23296:20140102:075947.513 resuming Zabbix agent checks on host "asterisk.p98.belkam.com": connection restored

as you can see, snmp started far earlier.

Thank you!

full log with debug 4

full log with debug 4

full log with debug 4

sorry, uploaded several times- got java script error for some reason

and how many unreachable pollers were there ? maybe this was a duplicate of ZBXNEXT-2359 ?

Hello!

Don't remember how much pollers were there,

Now, on 3.0.2 we have
StartPollersUnreachable=20

Now there are no such events for agent's devices, but I still have this errors for some very old devices using snmp:

19155:20160425:102417.012 SNMP agent item "ifInOctets41" on host "p98a-cc6006-1.p98.belkam.com" failed: first network error, wait for 15 seconds
19180:20160425:102510.648 resuming SNMP agent checks on host "p98a-cc6006-1.p98.belkam.com": connection restored

I guess that they are not fast enough in response.

I'm very sorry for asking this question, but are there any snmp timeout options for zabbix server?

Yes, since ZBX-4393 the Timeout configuration option is respected for SNMP checks. Regarding a more fine-grained timeout and retry configuration for SNMP checks, please see ZBXNEXT-1096.

also haven't seen any graphs on the busy pollers & unreachable pollers. were they all busy, maybe ?
overall this seems to be more of a support case. i'd suggest closing it as a "can't reproduce" case and continuing this on irc and other channels.

Aleksandrs, thank you for pointing to snmp timeout settings, I'll try this.
Anyway, I don't have these zabbix agent related problems at current 3.0.2 zabbix version.
So, looks like, problem is fixed, so, please, close this issue.

richiv, if you don't see any graph- this is because you didn't ask for them

Thanks for info! Closing as "Cannot Reproduce" then.