It took me a while to find all the little pieces to put together a custom rule that used data collected through TSQL. In fact, because it took so much work to pull all the little pieces together, I wrote an article describing how I did it. This isn't a wonderful new invention. It's just the publication of a bunch of research. If you're an Operations Manager expert, this isn't for you. If you're like I am, trying to find your way through the forest, this little bit of map will hopefully provide some assistance. Available at SQL Server Central.
I've spent five days struggling with this, but I finally have a custom performance collection up and running. The documentation available for Operations Manager 2007 (SCOM), especially for customization, especially for databases, is weak, and that's being kind. However, there are some guys working hard out there to make a difference. If you need help, the System Center Forum is the place to go. In particular, this little ditty on using property bag objects finally cracked the code for me. The good news, I'm going to write this up in an article for Steve over at SQL Server Central. The better news, I can finally report progress in my daily stand-ups. Sheesh!
It took a while, but I finally tracked down the information I needed. SCOM has an SDK. In the SDK is a nice little page showing how to test & debug scripts. One of the links on that page leads out here to a great article on the VBScript debugger. That's what I needed (along with the DLL's to register). Anyway, I've almost got my first monitoring script up & running. I'll post it on here once I do.
I had posted a problem I was having with SCOM below. I got a good answer to the problem in the Microsoft discussion group. It was a silly little thing. The "SQL" alias inside the object needed a reference placed within the Management Pack definition. I've still go so much to learn about this tool.
I'm stuck. I posted this last week on the Microsoft SCOM newsgroup, but I'm not getting any responses. On the off chance that the 5 people a day who read this site might know the answer, here's the question as posted: I'm sure this is something fundamental and simple, but I'm honestly stuck. I'm attempting to capture a new Performance Counter, Plan Cache, Cache Hit Ratio, Object Plans & SQL Plans. When I lay it out to create it, I get the attached error. I'm assuming based on the "Unknown alias" and the fact that I'm in a custom management pack, something is amiss with the Management Pack. What? Any help at all would be appreciated. Date: 3/28/2008 11:29:55 AM Application: System Center Operations Manager 2007 Application Version: 6.0.6246.0 Severity:…
I got a new SCOM book in the mail yesterday, System Center Operations Manager Unleashed. It's friggin' huge. I've just started reading through it and it looks pretty good. So far, it's much more thorough than the only other book available for SCOM, Mastering System Center Operations Manager 2007. I'm going to ready through more of it before I post a review to Amazon. I think I'll write up a review for the PASS book reviews too. Hey, if you have time on your hands, get on over there, request a book and write up your own review.
A friend has had an article posted over on SQL Server Central that's worth a read. Scott Abrants outlines the type of code that SCOM uses for SQL Server auditing and how you can leverage that outside of SCOM for other work you may be doing. As noted in the comments, it's also worth taking a look at how Microsoft set up it's rules and monitors as a method for setting up your own.
I'm working through implementing SCOM, specifically the SQL Server Management Packs. I've been tweaking and tuning and for the most part I'm really impressed with how MS configured the defaults. Until today. Since I've hit all the really big and important monitors & rules (at least the ones that have caused problems) I'm starting to drill down a bit. I just hit the Duration monitor for SQL Agent Jobs. The run time for a warning state is 1 minute and an error state in 2 minutes. OK. So any backup on a database over about 2gb in size is going to go into a warning state? I don't think so. What the heck were these guys thinking of?