[1.4] ZBX_TCP_READ() failed [Interrupted system call]

[globe]

HI @all,

my zabbix 1.4 server often dies with this message:

3311:20070627:133916 Get value from agent failed. Error: ZBX_TCP_READ() failed [Interrupted system call]
  3311:20070627:133916 Host [lx080]: first network error, wait for 15 seconds
  3384:20070627:133919 Timeout while answering request
  3384:20070627:133919 Get value from agent failed. Error: ZBX_TCP_READ() failed [Interrupted system call]

does anybody have the same failure and knows how to solve it?

Thanks Globe

[starstream]

We just upgraded to 1.4 and notice a similar problem.

Our zabbix_server did not die but we have seeen lots of thos messages in the log.

3153:20070628:122637 Timeout while answering request
3153:20070628:122637 Get value from agent failed. Error: ZBX_TCP_READ() failed [Interrupted system call]
3153:20070628:122637 Host [ZABBIX Server] will be checked after 60 seconds

I have just found out that if you disable all net.tcp.service items it disapears.

This is from the agentd log:

3224:20070628:122647 TCP expect error: Cannot connect to [127.0.0.1:143] [Operation timed out]
3224:20070628:122647 Sending back [0]
3224:20070628:122647 Process listener error: Cannot connect to [127.0.0.1:143] [Operation timed out]
3224:20070628:122647 Processing request.
3224:20070628:122647 In check_security()


My guess is that in the 1.4 upgrade the socket timeout is higher than before or at least higher that zabbix_server is waiting for an answer.

We are using

FreeBSD 6.2-RELEASE FreeBSD 6.2-RELEASE #0: Fri Jan 12 08:43:30 UTC 2007 [email protected]:/usr/obj/usr/src/sys/SMP amd64


I will continue to investigate......

[and.din]

i think you have to check th zabbix_agentd.conf
It looks like you inserted the wrong server-ip.

[starstream]

The agent and the server are on the same server so everything is OK.

I did hoverver find a solution.

All machines that i get timeouts from use FreeBSD's SPF firewall.

And if you have the following parameter set

sysctl net.inet.tcp.blackhole

The server does not respond with a PORT closed if trying to open a port that is closed. This will make it harder to portscan the server and to find out that it exists.

Setting this parameter to 0

sysctl net.inet.tcp.blackhole=0

Removes all timeout issues.

This is a problem only when probing ports that are not listening and if you have blackhole enabled. Still do not know what path to choose (disable blackhole, set the timeout higher, find another way of monitoring ports)

Cheers!

[zerg]

The agent and the server are on the same server so everything is OK.

I did hoverver find a solution.

All machines that i get timeouts from use FreeBSD's SPF firewall.

And if you have the following parameter set

sysctl net.inet.tcp.blackhole

The server does not respond with a PORT closed if trying to open a port that is closed. This will make it harder to portscan the server and to find out that it exists.

Setting this parameter to 0

sysctl net.inet.tcp.blackhole=0

Removes all timeout issues.

This is a problem only when probing ports that are not listening and if you have blackhole enabled. Still do not know what path to choose (disable blackhole, set the timeout higher, find another way of monitoring ports)

Cheers!

I have simular problem. When i check a lot of connections to my TCP port.

Have you solved problem?

uname -a
FreeBSD localhost 6.1-RELEASE-p10 FreeBSD 6.1-RELEASE-p10 #0: Wed Nov 15 13:53:18 UTC 2006

[starstream]

Hi!
Yes, i solved my problem. The problem only existed when i tried to check ports that where closed and when i had blackhole enabled.

My solution where to remove all portcheck for services i know dont listen on the machine with blackhole enabled.

Cheers!