Recovery action to triggers ?

[JohanLanderholm]

Hi,

I'm quite new to zabbix, has been running it for some basic monitoring requirements only notifying me on failures in web monitoring and such.
It works very good.
But - now I wish to take the "next step", and use it to automatically salvage some normal issues in my (mostly) linux environment.
Such as a failing process or service on linux.
I have been working with BMC Patrol some time ago and there was something called "Recovery Action" that could be triggered from a threashold trigger. So when a (tmp) disk was full, the agent could erase temporary files, or perhaps truncate log files to make room for the system to continue working reliably. Without any notification, unless the recovery action failed.
I would like to use the same kind of function in the zabbix agent.
I have searched the forum and I don't see any questions on this topic. Which makes me think - maybe the zabbix agent can't do any recovery scripts automatically, from a trigger event?

Is it possible to do ?

Best regards,
Johan

[cyber]

There is this thing... called Manual.. It has something about recovery operations and remote commands...

https://www.zabbix.com/documentation/current/en/manual/config/notifications/action/recovery_operations

https://www.zabbix.com/documentation/current/en/manual/config/notifications/action/operation/remote_command

It does work abit differently than BMC... (that awfulness still brings shivers to my spine..)

[JohanLanderholm]

Thanks, the manual was my first try.
It talks about Recovery operations - and it shows how to setup notifications. This is not my goal.
Remote commands seems to be the thing, but it need to be locally setup in the agent node. Of course - you need to edit the sudoers file, and define the script in the global script menu.
This feels like something that is not used very often in zabbix?
The BMC Patrol was a very good monitoring product that was scalable and supported most of the platforms at that time.
It was very focussed on keeping system up and running. Instead of notifying administrators of failures, to be fixed after the failure.

[cyber]

Thanks, the manual was my first try.
It talks about Recovery operations - and it shows how to setup notifications. This is not my goal.

Not just notifications... Remote commands are one type of "notifications"... execute on host, etc..

Remote commands seems to be the thing, but it need to be locally setup in the agent node. Of course - you need to edit the sudoers file, and define the script in the global script menu.
This feels like something that is not used very often in zabbix?

These are still triggered by server, while triggers are firing. As agent just collects and does not make decisions, then you cannot have actions executed directly by agent, Decision point is elsewhere...
I have not used them TBH, these were not available for a quite long time at all .. so maybe the usage is a bit smaller...

The BMC Patrol was a very good monitoring product that was scalable and supported most of the platforms at that time.
It was very focussed on keeping system up and running. Instead of notifying administrators of failures, to be fixed after the failure.

You are not going to convince me.. It was nightmare to configure and maintain... and the cost of it... All these km-s (with last change in 1995 .. ), that did not work as any normal human being would expect them to.. nono... lets not go there again..

[tim.mooney]

You sound traumatized Cyber! BMC can't hurt you any more. :-)

[JohanLanderholm]

Not just notifications... Remote commands are one type of "notifications"... execute on host, etc..
These are still triggered by server, while triggers are firing. As agent just collects and does not make decisions, then you cannot have actions executed directly by agent, Decision point is elsewhere...
I have not used them TBH, these were not available for a quite long time at all .. so maybe the usage is a bit smaller...

You are not going to convince me.. It was nightmare to configure and maintain... and the cost of it... All these km-s (with last change in 1995 .. ), that did not work as any normal human being would expect them to.. nono... lets not go there again..

I'm not trying to. But I would really like to learn how to perform actions in response to threshold alerts in Zabbix. I think most of the people want to be notified by zabbix when something goes wrong. I would like zabbix to remedy the situation *before* it become a real problem. I think many recovery scripts is quite specialized to the node/os that it should be run on.
I have found out you can create recovery scripts and how to fire them from a trigger. It was well hidden, but I found it. It isn't as strait forward as I expected it.

Don't get me wrong - I enjoy the zabbix system. I have been working with Nagios, CheckMK, OP5, HP OpenView, IBM Tivoli and initially BMC Patrol. I think the patrol agent had some features I really miss today, like the event handling, the local parameter database, the possibility to configure the agent to be "quiet" on the network, unless you needed some parameters to be transferred centrally.
Today - I'm trying to convince a checkMK shop to switch to zabbix, but they have a lot of specialized custom scripts for tons of custom written applications. It's not simple to switch. But I think they have to in the long run.