Does anyone have a script for doing bulk configuration of "Dynamic Threshold" monitors (i.e. qos_engine)?
For example, applying "dynamic thresholds" to the TOTAL target for all sources of QOS_CPU_USAGE.
I don't know if you'll be able to easily accomplish this via our QoS Monitor option in the SLM, sadly, that feature doesn't scale well at all, it really feels like more of an 'RFI checkbox' than a usable feature since it's so tedious to implement in bulk... Also, if I recall correctly I believe it's limited to only 100-150 monitors.
Check out this thread for details on how you might be able to hack it with SQL:
A better solution for this will probably be the script Jim wrote for shop.com a few years back that accomplishes something similar using a custom probe he wrote in perl.
Good luck and please let us know how you end up resolving this!
Thanks for the reply - a couple of questions for you:
1) There is a thread in the forum with a subject of "Threshold Adviser" for which I think you attached Jim's code as referenced in your comment below. I may be mistaken but this solution also looks like it only monitors single instances of a metric per definition. Jim, if you're listening, do you have a version which supports wildcards or regex?
2) When you say that qos_engine is not scalable I assume you are referring to performance rather than the pain of configuring it. Is the 100-150 limit based on a controlled benchmark test? Do you have any more detail?
Hi Ray. You are correct that the probe I wrote monitors single instances of QoS metrics. Each profile requires a QoS name, source and target as the key values. Shop.com had some very specific values that it wanted to monitor in this way. Also this probe was a predecessor to the qos_engine - QOS monitors solution from Oslo.
Q. When you say that qos_engine is not scalable I assume you are referring to performance rather than the pain of configuring it. Is the 100-150 limit based on a controlled benchmark test? Do you have any more detail?
A. I guess I would be referring to both, assuming there is a limitation of ~150 monitors. I'm basing this off something Carstein said when he introduced this feature in New Orleans a couple of years ago, I may be remembering wrong and/or it may have been increased.
My real issue with it not scaling well is because there is no easy way to implement it for several servers, targets, etc.
Retrieving data ...