Zabbix host loss of DNS entry

[Justin417]

Hello!

I'm chasing down a rather odd problem I'm having with a couple agents in my environment. For context, we have 1000+ hosts so far monitored without issue, but 2-3 nodes have seen this issue multiple times.

Essentially, this agent's DNS entry disappears, the server no longer monitors the agent, and our slack alerts begin to fail (Error message: "Slack notification failed : Field "host_conn" cannot be empty"). Viewing the host in Configuration > Hosts, the DNS name appears blank, and the IP address reverts to 127.0.0.1:





The issue can be resolved by manually re-adding the DNS entry and clicking "Update" - However the issue randomly re-occurs (sometimes weeks, sometimes months from when manually corrected).

We do utilize autoregistration from the agent side, however we use the same template across all our servers and have not seen this issue since. For context, here's a truncated version of our agent configuration file (commented lines removed):

PidFile=/run/zabbix/zabbix_agentd.pid
LogFile=/var/log/zabbix/zabbix_agentd.log
LogFileSize=0
Server=zbxprox1.ams1.ourdomain.com
ServerActive=zbxprox1.ams1.ourdomain.com
Hostname=mtlctrl1.ams1.ourdomain.com
HostnameItem=mtlctrl1.ams1.ourdomain.com
HostMetadata=AMS1-Linux-Server
Include=/etc/zabbix/zabbix_agentd.d/*.conf
Include=/etc/zabbix/zabbix_agentd.conf.d/*.conf
TLSConnect=psk
TLSAccept=psk
TLSPSKIdentity=mtlctrl1.ams1.ourdomain.com
TLSPSKFile=/etc/zabbix/zabbix_agentd.psk

Any help would be greatly appreciated to try and figure out what's causing the DNS entry of our hosts to disappear.

Thanks!

[LenR]

Are these by chance windows hosts?

[Justin417]

Hello!

No, these hosts are all Ubuntu 20.04 LTS utilizing the standard zabbix agent, version 5.4.

Thanks! Have you seen similar on Windows?

[LenR]

Yes, windows will sometimes not fill in hostnameitem or hostmetadataitem, I think the agent makes some windows API call that returns garbage. It's very sporadic.... I just excluded the host in the autoregistration action so it doesn't corrupt that host config and it's no longer a problem. It seems to be older, out of date windows servers.

Do you use HostName or HostNameItem? On Linux, since we use Puppet to deploy the agents, we just set HostName, but Windows wants to use some powershell script to get hostname, which I think causes the problem, but they worship at the grave of PowerShell.....

[Justin417]

LenR,

Thank you for the suggestion, I suppose just disabling the autoregistration works, we've done so and so far the issue hasn't re-occurred for this host. Very odd.

Thanks!