Another thread for nasty UNREACHABLE-problem

[jgerry]

We're also having tons of problems with "Host is Unavailable" problems with triggers. For the same host, some triggers are fine, some have the "Host is Unavailable" listed beneath them.

Nothing we do seems to fix the issue once it starts. For some checks, the data collection just seems to slow down and stop. We'll see checks, say, every 10 minutes for months, then we'll start seeing a few small gaps in data collection, then huge gaps, then no data at all.

Server version 1.6.6, agents mostly now upgraded to 1.6.6, some still 1.6.4.

[jpriceit]

This is a problem with us, too. We are running v1.6.6 with mixed versions of agents including v1.6.6, v1.6.4, and v1.4. We are not using dependencies, either.

Would I be able to downgrade to v1.6.4? Were there any DB structure changes between those versions or any other reason why I would not be able to do so?

[richlv]

database should be compatible in all 1.6.* versions

[jpriceit]

database should be compatible in all 1.6.* versions

Great! Thanks.

[elvar]

This is a problem with us, too. We are running v1.6.6 with mixed versions of agents including v1.6.6, v1.6.4, and v1.4. We are not using dependencies, either.

Would I be able to downgrade to v1.6.4? Were there any DB structure changes between those versions or any other reason why I would not be able to do so?

Have you noticed any of these problems in 1.8?

[dougnaka]

Still happens in 1.8

I tried Zabbix 1.6.2, 1.6.5, 1.6.6, and finally 1.8 and this exact thing happens randomly on lots of my hosts (75 total).

I have finally given up trying to fix it, as it would also not accurately tell me if hosts were down. I disabled the trigger "Server Template_Linux is unreachable" (also for all templates in use.

I then added a trigger on agent.ping. I don't recall if I had added the basic one first, or just changed it, but the net result is I have this trigger.

Name : {HOSTNAME} Ping Status
Expression : ({Template_Linux:agent.ping.last(0)}#1)|({Template _Linux:agent.ping.nodata(120)}=1)

Severity : High


This seems to work well, I have timed it and it's just over 2 minutes from when I disable an agent to when this alerts me.

[LenR]

I'm at 1.6.7 and have 2 time servers that go unreachable several times a day for 1 minute. I tried the agent.ping test and it does the same thing. 100+ other servers are OK.

[LenR]

OK, more news. I've been able to reproduce this problem with ping, so it's NOT ZABBIX! It really looks like a very intermittent network problem on a few hosts that provide a particular service.

[gullevek]

Same problem with 1.8.1

I have the same Problem with 1.8.1. Sometimes a host is marked unreachable, but all the data is still collected without a problem.

The host is up, but marked unreachable. I really would love to know why.

[LenR]

After my network problem, I still had other problems. I changes from 5 to 40 trappers on the server and that seems to have helped a lot.

My zabbix server is one of the few remaining physical systems and I had a run of hardware problems. After a daytime restart, I had lots of things queuing, that's how I happened on to this.

This is still 1.6.7.

[Surge]

Problem still exists in 1.8.3

I can confirm that this problem still exists in version 1.8.3.

I'm monitoring hosts behind bad connections (regional offices) and sometimes the connections go down for several hours.
When the connections come back up Zabbix starts collecting new data but triggers dependent on the "status" key get stuck in their previous state.
The other items are being updated for the host but the status item keeps getting "no data" so the item remains "UNKNOWN".
I have to delete and recreate the host to fix the problem.
I could use ping but that also has limitations because some firewalls drop or limit ICMP traffic.

Bug report 978 best describes the problem https://support.zabbix.com/browse/ZBX-978 but maybe this is related to 2632 as well https://support.zabbix.com/browse/ZBX-2632.