My 2cents on backing up performance data … and general best practices for saving performance collections.
First, I agree with Jim that if the management collection objects are causing issues wth your backup, exclude them.
However, you do need to have a plan for saving performance data from key timeframes.
I recommend keeping 8 days of management collections online for comparison purposes. You want to be able to compare back one week if something goes wrong.
And for many businesses I’ve worked with, you want to be able to compare Monday to Monday, Tuesday to Tuesday, etc., as workloads do vary throughout the week.
You should identify critical timeframes - the busiest day of the week, week-end/month-end/year-end processing and keep those collections online and also ensure you have backups of them.
To keep the collection objects online, simply move them to a different library and Collection Services will not expire them. When moving performance data around, use the CPY/MOVPFRCOL commands.
Or you could change the expiration date on the *MGTCOL object to be *NONE.
My preference is to have a separate library for backing up performance collections. You keep the *MGTCOL objects in there, and optionally create the data into the files when you need to review the data.
Periodically review the collections you have saved, and replace old collections with newer ones over time. You could easily write a little CL program that runs on Sunday morning to copy/move the important collections from the prior week into your library and deletes the oldest version. If I did something like this, I’l probably keep a rolling window of 3 to 4 collection objects.
You should back up the collection object from your busiest day of the week. If you have a different day that does some business-critical processing, you should also back up that collection.
These collections may be useful, again if something goes wrong, to have baseline data to compare against.
Keeping the management collection objects is more important than keeping the data in the Db2 files - you can always recreate your file data from the *MGTCOL objects.
Even if you don’t think you have any performance concerns, you should run Collection Services and keep collections for baseline data.
Collection Services collects a LOT of information - more just performance data - and that additional information could be useful.
Also, if you are on 7.3 and have not turned on Historical Data Collection, do so. It’ll give you trends over time.
If you’re interested in these topics and will be at COMMON, I have several performance sessions on the agenda.
Dawn
Email: dawnmayican@xxxxxxxxxxx<mailto:dawnmayican@xxxxxxxxxxx>
Web: Dawnmayi.com<
http://Dawnmayi.com>
Twitter: DawnMayiCan
Blog:
http://ibmsystemsmag.com/blogs/i-can/
LinkedIn:
https://www.linkedin.com/in/dawnmay/
midrange.com
