ZBX_TCP_READ() failed [Interrupted system call]

[mlange]

We got this error when monitoring a Java VM using Zapcat. Unfortunately the host can't be restored since the error occured. We see a lot of network connections being in CLOSE_WAIT state:

tcp6 52 0 ::ffff:IP.:14919 ::ffff:IP_ZABBIX:59364 CLOSE_WAIT
tcp6 52 0 ::ffff:IP.:14919 ::ffff:IP_ZABBIX:47332 CLOSE_WAIT
tcp6 52 0 ::ffff:IP.:14919 ::ffff:IP_ZABBIX:45284 CLOSE_WAIT

Around 2000 connections are hanging and they are not released. The agent seems to refuse new connections. What could be the cause for this, why are the connections not closed?

Thanks,
Marek

[Kees Jan Koster]

Dear Marek,

I have no idea why they would not be closed. I have never tested Zapcat with ipv6. Does this problem also happen if you limit the VM to ipv4?

Kees Jan

[mlange]

Good question. Since the problem occured on a staging environment I cannot test this with IPv4. It has never occured again since we restarted the server (and thus the Zapcat agent). All connections (and open files) were released after the restart. But I would like to know why this happened and what could be done to avoid this later in production. Thanks for your answer.

[Kees Jan Koster]

Dear mlange,

Hmm. I'm afraid I don't know. The code I have written is quite straightforward. You could root through it to see if I made any mistakes.

Other than that, please let me know if this happens again.

Kees Jan

[mlange]

It has not happened again yet. However I am experiencing a similiar strange problem: for one host using zapcat a trigger regularly gets true although the JVM looks ok. I have this expression:

{JVM_TEMPLATE:jmx[java.lang:type=Runtime][Uptime].nodata(240)}=1

It gives me an alert: ALERT: JVM unreachable for 240s: High

Do you have an idea why this errorneous alert is generated?