Can't run agent - Cannot create IPC key for agent collector

[devboy]

Hi,

I've recently installed Zabbix Server and several agents on our production machines. Today I tried to install the agents on a batch of VMs that we have. These VMs are of a different configuration than the production machines and are running CentOS 5.5 (production are running CentOS 5.4).

When I get to the point I need to start zabbix_agentd it fails. In the log I find this batch of messages:

zabbix_agentd [9297]: Can not create IPC key for path '/etc/zabbix/zabbix_agentd.conf', try to create for path '.' [Permission denied]
  9297:20101031:142428.035 Zabbix Agent started. Zabbix 1.8.3 (revision 13928).
zabbix_agentd [9297]: Cannot create IPC key for path [/etc/zabbix/zabbix_agentd.conf] id [l] error [Permission denied]
zabbix_agentd [9297]: Cannot create IPC key for agent collector
zabbix_agentd [9997]: Can not create IPC key for path '/etc/zabbix/zabbix_agentd.conf', try to create for path '.' [Permission denied]
  9997:20101031:142853.791 Zabbix Agent started. Zabbix 1.8.3 (revision 13928).
zabbix_agentd [9997]: Cannot create IPC key for path [/etc/zabbix/zabbix_agentd.conf] id [l] error [Permission denied]
zabbix_agentd [9997]: Cannot create IPC key for agent collector

I googled and found regarding stuck semaphores and removed them using ipcrm, but next time I try and run it does the same thing.

Any idea where this could be coming from? If it's permissions, then permissions on what exactly?

Cheers,
DB

[devboy]

Solved

Took me 4 hours but figured it out eventually...
I was missing permissions on the /etc/zabbix/ where zabbix_agentd.conf resides.

Did 'chmod -R 755 /etc/zabbix/' and solved it.

[walterheck]

Hmm, took me a while to figure out as well, going through the same process. Pretty annoying. Are you using the RPM's from http://repo.andrewfarley.com as well?

[onefang]

I'm getting the same issue, two of the production servers are reporting that error, the rest are fine. A new install of 1.8.4. All servers had /etc/zabbix copied from the original one I compiled things on. I tried that chmod, but I still get the same errors.

So, what else should I try?

[devboy]

I've used this to install the server the first time. Now that I migrated our production I did it from source to not have to disable selinux on the Zabbix Server box (the script demands it for some reason).

For the client I always installed from source and customized some script for myself.

Here's the procedure I'm following installing an agent on a vanilla CentOS 5.5. (you might need yum install gcc as well)

Download, extract and cd into your Zabbix release (verified on 1.8.3 and 1.8.4) and do the following:

sudo ./configure --enable-agent --prefix=/usr/local/zabbix
sudo make install 
sudo /usr/sbin/adduser zabbix
sudo mkdir /etc/zabbix
sudo cp misc/conf/zabbix_agentd.conf /etc/zabbix
sudo vi /etc/zabbix/zabbix_agentd.conf
	Server=your.server.ip
sudo vi /etc/init.d/zabbix_agentd
	Paste init.d script from below
sudo chmod 755 /etc/init.d/zabbix_agentd
sudo chmod -R 755 /etc/zabbix/
sudo chkconfig --level 345 zabbix_agentd on
sudo /sbin/service zabbix_agentd start

The init.d/zabbix_agentd script:

#!/bin/bash
#
# chkconfig: - 55 45
# description: zabbix_agentd
# probe: false

# Source function library.
. /etc/rc.d/init.d/functions

# Source networking configuration.
. /etc/sysconfig/network

# Check that networking is up. If you are running without a network, comment this out.
[ "${NETWORKING}" = "no" ] && exit 0

RETVAL=0
progdir="/usr/local/zabbix/sbin/"
prog="zabbix_agentd"

start() {
        # Start daemons.
        if [ -n "`/sbin/pidof $prog`" ]; then
                echo -n "$prog: already running"
                failure $"$prog start"
                echo
                return 1
        fi
        echo -n $"Starting $prog: "
        # we can't seem to use daemon here - emulate its functionality
        su -c $progdir$prog - $USER
        RETVAL=$?
        usleep 100000
        if [ -z "`/sbin/pidof $progdir$prog`" ]; then
                # The child processes have died after fork()ing, e.g.
                # because of a broken config file
                RETVAL=1
        fi
        [ $RETVAL -ne 0 ] && failure $"$prog startup"
        [ $RETVAL -eq 0 ] && touch /var/lock/subsys/$prog && success $"$prog startup"
        echo
        return $RETVAL
}
stop() {
        RETVAL=0
        pid=
        # Stop daemons.
        echo -n $"Stopping $prog: "
        pid=`/sbin/pidof -s $prog`
        if [ -n "$pid" ]; then
         kill -TERM $pid
        else
         failure $"$prog stop"
         echo
         return 1
        fi
        RETVAL=$?
        [ $RETVAL -ne 0 ] && failure $"$prog stop"
        [ $RETVAL -eq 0 ] && rm -f /var/lock/subsys/$prog && success $"$prog stop"
        echo
        return $RETVAL
}
restart() {
        stop
# wait for forked daemons to die
        usleep 1000000
        start
}

# See how we were called.
case "$1" in
        start)
                start
                ;;
        stop)
                stop
                ;;
        restart)
                restart
                ;;
        condrestart)
                [ -f /var/lock/subsys/$prog ] && restart
                ;;
        *)
                echo $"Usage: $0 {start|stop|restart|condrestart}"
                exit 1
esac

exit $?

Disclaimer: I'm no linux guru. I might have done awfully stupid things and will have no idea! Having said that, I'm serving half a million unique users a day with boxes with this setup so...