Last week I discovered a problem with the Dependency Rollup monitors inside one of my management packs. After a re-import of the management pack the dependency rollup monitors didn’t work anymore. The monitors would not initialize after the re-import. When I updated the management pack anything was fine but with a re-import of the Management Pack I ended with broken dependency roll-up monitors.
Let’s start at the beginning and explain a little bit more about the components which are in the Management Pack. The Management Pack consists of the following components:
- A Dynamic Group with OleDB transactions;
- A Dependency Rollup monitor which give the dynamic group a combined health state of the transactions;
- A Distributed application with one component group which has one member: the dynamic group from above;
- A Dependency Rollup monitor which gives the component group the health state of the dynamic group;
- A Dependency Rollup monitor which gives the distributed application the health state of the component group.
When I import the management pack for the first time everything is working fine. All objects gets a health state and all monitors are working. When I update this management pack with a new version everything still works without any problems. But the problems are there when I delete the management pack from the SCOM environment and import it again with the same name and version after some time (couple of minutes or couple of hours). The Dependency Rollup Monitors of the Distributed Application and the Distributed Application component will not work anymore. I tried the following to resolve the problem:
- Force a re-initialize of the monitor by putting it temporary in Maintenance Mode
- Renamed the ID and/or name of the monitors and then re-imported the Management Pack
- Increased the version of the Management Pack and then re-imported the Management Pack
All of the above didn’t give me a working Management Pack, in the end I have found the following workarounds for this problem:
- Delete the management pack, clear the cache on all Management Servers and then import the management pack.
- Delete the management pack, change the name of the management pack ( I tested this with adding a timestamp to the name of the MP) and import the new one.
With the above workarounds I got the MP working again, but still this is not a very nice workaround. The first workaround is not a acceptable solution and the second workaround is also not the ‘nicest’ workaround. So I decided the log a support case at Microsoft for this issue. Currently I’m in the process of investigating the issue with a Microsoft Support Engineer.
If these problems sounds familiar to you please leave a message. I will update this post when more information is available!