zabbix v1.1.4 - stops monitoring client

[dcrandall]

Hi,

Zabbix_server v1.1.4

I'm having this problem where if I reboot a client machine, when it comes back, zabbix_server does not start monitoring it again.

I can kick start the process if I do this:
mysql -u root -D zabbix -e 'update hosts set errors_from=0'
Then it start monitoring again.

Can anybody shed light on this? I'd like to know what is happening? It's a pain because sometimes machines need to be restarted, and they have to call me everytime to get zabbix montitoring to work.

Thanks,
Daniel

[Alexei]

ZABBIX does not start monitoring of unreachable hosts immediately but after 60 (or 120?) seconds by default.

[DiedX]

60. Adjustable BTW.

[dcrandall]

I am aware of the unreachable time limit and have verified that it is 60 seconds on my server.
The thing is, that my clients go unreachable and then come back and zabbix_server never updates. They just sit there with nodata untill I run the command:
mysql -u root -D zabbix -e 'update hosts set errors_from=0'

It's pretty annoying because sometimes under heavy load, a host will lag in responding, so zabbix changes to it to unreachable and never changes back when the host is more responsive.

I never had this problem until upgrading to v1.1.4.
The server and hosts are all FreeBSD 5.x or 6.x.
As I have described in other posts regarding zabbix_agentd, it seems that support for FreeBSD has taken a hit in zabbix 1.1.x stable.

Thanks for your help,
Daniel.

[pdwalker]

What do your log files say?

Can you turn on debug logging and learn anything new about why the server is not connecting to the client any longer?

With my installation, the server has not failed to restart monitoring a cliient after it has been "away" and marked as unavailable.

- Paul

[dcrandall]

Hi,

Thanks for your reply.

For instance, I have a host that is responsive, returning values from zabbix_agentd -p, and also from the server side with zabbix_get.

In the log I have:
075050:20061231:032951 Host [contb004]: first network error, wait for 15 seconds

That message is obviously from 2 days ago and the host is still showing [no data] in the graphs.

If I run the mysql command in my previous post, it will come back.

I really believe that the problems I'm having are specific to FreeBSD 5.x and 6.x with zabbix 1.1 through 1.1.4

Daniel

[dcrandall]

This is getting serious

Hi,

Last night a client host dropped out the same way I described above. This is what was left in the zabbix_server.log...

075054:20070104:031059 Host [contb002]: first network error, wait for 15 seconds
075049:20070104:031102 Host [contb002]: first network error, wait for 15 seconds
075050:20070104:031102 Timeout while receiving data from [contb002]
075050:20070104:031102 Getting value of [betacontent.photoshow.com] from host [contb002] failed
075051:20070104:031102 Host [contb002]: first network error, wait for 15 seconds
075052:20070104:031111 Timeout while receiving data from [contb002]
075052:20070104:031111 Getting value of [vm.memory.size[free]] from host [contb002] failed
075053:20070104:031111 Timeout while receiving data from [contb002]
075053:20070104:031111 Getting value of [vm.memory.size[shared]] from host [contb002] failed
075050:20070104:031132 Timeout while receiving data from [contb002]
075050:20070104:031132 Getting value of [system.cpu.load[,avg1]] from host [contb002] failed

From that moment on the server reported [no data] from that host.

Unfortunately, later on, that host crashed.

Because it was in [no data] state, we didn't receive any alerts. Meaning nobody was paged, meaning that the CEO of the company called first thing this morning to tell us that our service was not working.
I don't have to tell you that this is not good.

Now I need to find some resolution to this problem, or it's quite likely that my management will lose confidence in Zabbix.
This is not the only host that is having this problem, it happens to all of my FreeBSD machines from time to time and I have to massage the database myself to get the server to start monitoring them again.

The server is Zabbix v1.1.4 running on FreeBSD 5.4, and the clients are all either FreeBSD 5.x or 6.x. As I've mentioned before, this is one of a few major problems with FreeBSD support in Zabbix 1.1.x

Thanks

Daniel

[Mintairov Mihail]

I have got the same problem with zabbix. I am monitoring a wifi routers by using SNMP-items
and sometimes zabbix tells that some items have [no data] state and ERROR_FROM
value is not 0, however the wifi routers work correctly, and if I try to get some
values with SNMPGET command everithing works fine.

I have made the test with a network sniffers to see if zabbix tries to define a state of "not working"
hosts, and I saw that zabbix did nothing, so if ERROR_FROM is not 0 then zabbix does nothing with those hosts.

[Alexei]

I have got the same problem with zabbix. I am monitoring a wifi routers by using SNMP-items
and sometimes zabbix tells that some items have [no data] state and ERROR_FROM
value is not 0, however the wifi routers work correctly, and if I try to get some
values with SNMPGET command everithing works fine.

Are you on FreeBSD as well?

[Mintairov Mihail]

Are you on FreeBSD as well?

Yes I use FreeBSD 6.0-STABLE