Remote command failing

[Mark Thornton]

I'm using Zabbix 2.2 on Ubuntu 14.04 and on our newest machines system.run[command] is failing when the command takes longer than echo hello to execute. I can run the command on the machine as 'zabbix' and it takes under a second, but if run via the zabbix agent it sits in the process list doing nothing until the timeout is reached.

The machines on which it fails are about 3 times faster than the fastest of those on which it works (i7-5820 vs i5-2400, 6 core + hyperthreading vs 4 core).

I have run out of ideas as to what do with this.

An example of a command which fails is
/usr/lib/update-notifier/apt-check

(reports number of packages which can be updated).

[Mark Thornton]

The simplest command which fails is
lsb_release -s -d

while
df -h
succeeds

For now my workaround is to create a cron job which writes the answers to a series of files. Zabbix agent then retrieves the content of the files with vfs.file.contents.

[bbrendon]

Maybe the agent is running as a user with bad permissions or environment?

[Mark Thornton]

A user or environment in which lsb_release fails?

It is running under the zabbix user account (as standard). I have not been able to identify anything unusual or different from the machines on which it works.