Interrupted system call !!

[azurit]

hi,

i'm still having problems with hosts availability:

3020:20080305:161214 Timeout while answering request
3020:20080305:161214 Get value from agent failed. Error: ZBX_TCP_READ() failed [Interrupted system call]
3020:20080305:161214 Host [XXX]: first network error, wait for 15 seconds
3020:20080305:161214 Parameter [agent.ping] will be checked after 120 seconds on host [XXX]

this is happening about once per week for random host (also for localhost), i also get e-mail saying 'host unreachable'. hosts are, of course, always up and working.

Debian Etch, Zabbix version 1.4.5 dev (downloaded at 02.03.2008), both server and client. the same thing was happening on 1.4.3 .

please response me Alexei, this is very serious problem for me (zabbix is useless in this state, it's generating LOTS of false positives). do you need any other info ?

[Alexei]

ZABBIX has failed to connect to the agent due to timeout, as simple as that. Why do you think this is ZABBIX fault? I'd like to see at least one evidence before any troubleshooting!

[azurit]

ZABBIX has failed to connect to the agent due to timeout, as simple as that. Why do you think this is ZABBIX fault? I'd like to see at least one evidence before any troubleshooting!

i can't tell it for sure of course, but everything other is working fine. how can i find out what could cause this ? or how can i prove that it's zabbix fault ?

[azurit]

Alexei, i'm doing 'agent.ping' every 30 second and i have 'unreachability trigger' set like this:
{host:agent.ping.nodata(90)}=1

so, if there are no data for 90 seconds (3 checks), host is considered as unreachable. now i don't understand why i'm getting 'host unrerachable' because of _one_ network error ? there should be also two more checks to be made before considering host as unreachable.

[Alexei]

Alexei, i'm doing 'agent.ping' every 30 second and i have 'unreachability trigger' set like this:
{host:agent.ping.nodata(90)}=1

so, if there are no data for 90 seconds (3 checks), host is considered as unreachable. now i don't understand why i'm getting 'host unrerachable' because of _one_ network error ? there should be also two more checks to be made before considering host as unreachable.

The nodata(90) means that there is no data coming for the agent.ping for ANY reason: high latency, performance problems on ZABBIX server side, failed checks, network problems, anything!

[azurit]

The nodata(90) means that there is no data coming for the agent.ping for ANY reason: high latency, performance problems on ZABBIX server side, failed checks, network problems, anything!

yes, but nodata for _3_ checks (cos i'm doing 1 check per 30 seconds). there is only one message telling about network problems in the logs, which probably means, one check was false (nodata). what about two other checks ? why it is immidiately considered as 'unreachable' ?

[Alexei]

yes, but nodata for _3_ checks (cos i'm doing 1 check per 30 seconds). there is only one message telling about network problems in the logs, which probably means, one check was false (nodata). what about two other checks ? why it is immidiately considered as 'unreachable' ?

One check per 30 second under ideal conditions. In case of timeout situation ZABBIX protects himself by disabling the host for 60 (if I recall correctly) seconds. You've got one timeout, this is enough to make all three checks fail.

[azurit]

One check per 30 second under ideal conditions. In case of timeout situation ZABBIX protects himself by disabling the host for 60 (if I recall correctly) seconds. You've got one timeout, this is enough to make all three checks fail.

so what do you suggest as a solution ? how should i check for availability ?