Setup is Check_MK RAW v1.6.0p17 - Distributed Monitoring. Added 2 x extra CMK servers to better distribute the load (total of 5: Master + 4 slaves). I am having difficulty migrating from one instance to another on hosts with parents (the majority). Tried migrating multiple ways: both host and parent - failed. Only the host - failed. Only the parent - failed. So, what is the proper way of successfully migrating them in mass? The only way I found so far is to do a single host at a time, edit and remove the parent setup. Apply the changes, then edit the host and change the “Monitored on site” setting, applying and then go back to re-scan parent for that host (already on the new instance). Doing that to hundreds of hosts is no joke. So I hope I’m doing something silly.
You are right, it can be painful to split a site with many parent child relations into two.
What i do in such cases, i assign a whole folder to a new instance and check at config activation if there are error messages regarding a parent child problem.
That’s the only thing what i would do if i want to move a huge number of hosts between instances.
Not sure I follow. Does it mean you don’t synchronize settings from master to all instances? They are all independent but under the same distribution?
No i do a normal master/slave setup and only not assign the “monitored on” to every host but complete folders.
Now i can change the complete folder to another instance and can have a look at activation time if there are broken parent / child relations i need to fix.
Do you still keep hosts and parents in separate folders? Folders like: instance1/parent, instance2/parent, instance3/parent, etc, where isntance1=folder for hosts and instance1/parent=folder for parents? Or both hosts and parents are in the same folder (i.e.: instance1)? Thanks
What I am still unsure about is that instance parents (instance1/parent) each receive their instance name (instance1/parent/gw-instance1-IP). So, when I move the parent to instance2, the host still contains the previous name of gw-instance1-IP. How to resolve that?
Short Answer No. I more try to keep hosts monitored on one site grouped under one folder responsible for this site.
Folder1 - monitored on siteA
- Subfolder 1 - inherit the settings
- Subfolder 2 - the same here
Folder2 - monitored on SiteB
- Subfolder3 - inherit the SiteB from his parent folder
Now i can move a complete Subfolder to another site and have only to correct the broken parent childs after activation.
Oh this comes from the automatic detection of parents in the meaning of routing information or?
This is normally not used in my systems only for some objects with a complex unknown WAN connection.
The name by itself makes for the monitoring no difference the only point what’s important is the parent setting on every object.
Understood. Thanks for explaining. Mostly, my parent/child relationship is for the tracert discovery due to the complex LAN/WAN setup. So I use the auto-discovery of parents for this reason.
Correct and I apologize for the confusion. I assumed parent/child on the auto-discovery of the network hops, not another parent/child relationship.
Here is where things get complicated (so far) when moving/distributing the load. I haven’t found an easy way to “uncouple” them easily besides editing individually.
Still, your folder distribution helps indeed, in cases when I do not have parents tied to the host.
I am still trying to find an easier way.