If the Zabbix server goes down ...

[DonnEdwards]

If my Zabbix server goes down or is disconnected from the network (i.e. power loss or network cable unplugged), what happens to all the machines running the Zabbix Agent? Does the data get queued, or is it lost? How long is the data stored by the Agent, if at all?

Would the machines running the Zabbix Agent become unstable after a while?

Just curious.
Donn

[nelsonab]

I saw your title and was reminded of an old joke, but with a new tiwst... "If a Zabbix server dies at night and nobody get's an alert, did it really die?"

Active items will queue some data in the agent
http://www.zabbix.com/documentation/.../zabbix_agentd
Has more information about the BufferSend and BufferSize variables.

Passive items will not be stored in the agent.

I'm not aware of any instability in the agent if the server is not up for extended periods of time.

[DonnEdwards]

Thanks for the reassurance. The last thing I need is a whole bunch of users on my back while I'm trying to fix the server.

The joke reminded me of a similar joke for teaching programming: "write a function which tells you if the power to the CPU is on". Of course the answer is trivial:

function IsPower() as boolean
   IsPower = True
end function

[tchjts1]

May be a good time to consider the use of a Zabbix Proxy server. If your Zabbix app server goes down, the proxy can retain information coming from the hosts.

You can set a variable in the proxy conf on how much data to retain. I personally have my proxies (We have 13) set as low as possible, at 1 hour.
if you start setting it higher, that's all the more information that the Zabbix app server has to crunch when it comes back on-line.

We lost our app server for awhile... 8 hours I think. At the time, I had the proxies set to retain 6 hours of data. When everything connected back up, it was taking the Zabbix App server one hour to process 2 hours of retained proxy info. So it would have taken 3 hours just to get the older info processed before it started even working on current data.

Needless to say I stopped the Proxy process and wiped the Proxy DB's clean, then reconnected.

[qix]

I have Debian running on Sparc, drop me a line if you're interested Alexei.

[ErikEngerd]

Hi,


In version 1.8.12 of zabbix agent, BufferSend also determines a minimum for how long data will be kept locally. I checked this by setting it to 15 minutes and I got data only every 15 minutes. I checked this using a trigger checking for data received in the last 5 minutes (nodata(300)), and this repeatedly got into PROBLEM and OK state.

It would be nice to have one configuration setting for the maximum time to buffer with perhaps a connection retry interval and also a minimal frequency with which the agent tries to contact the server. I definitely need this because if a monitored system has a network problem or extreme load problem I would like to get statistics afterwards also for the period that the server was basically inaccessible.

Cheers
Erik