Rules taking up alot of OpsDB Space

[maekee]

Hi,

We are using nWorks VMware SCOM MP 5.5.

After updating to version 5.5 these Performance Rules:
VMHost-Disk writes
VMHost-disk deviceWriteLatency
VMHost-disk totalReadLatency
VMHost-Disk aborts
VMHost-Disk reads
VMHost-Disk resets
VMHost-disk deviceReadLatency
VMHost-Disk kbRead
VMHost-Disk kbWritten
VMHost-disk totalWriteLatency

are taking up aloooot of space in OperationsManagerDB.
Each Counter takes about 640 MB.

At what Interval does these counters collect data?

I want to take this amount down, where is the best to do this?
There is no override for this on these Rules.

Help please

/Maekee

[Alec King]

Hi maekee,

Actually - all performance rules in 5.5 will take LESS space than they did in 5.0.
This is because in 5.,5. we
1. use Optimised Performance Providers in the MP - so each counter is only stored if the value is outside the deviation from previous sample
2. have configurable collection interval for every class - by default these metrics (from Host Disk Details) have an interval multiplier of 3x. Default interval is 5min, so a data point is sent only every 15 min.

You can reconfigure both of the above points, if you want to use less space in the Ops Mgr Db.
1. Optimised Providers can be overriden for every Rule - use Ops Mgr UI, go to Authoring\Rules, search or scope for 'nworks' rules, and override your desired rule (e.g. 'HOSTDISK.Collect.totalWriteLatency'). I suggest you increase Tolerance (how far the sample has to deviate from previous sample before it is recorded), and/or MaximumSampleSeparation (how often a data point will be stored anyway, whether it deviates +- Tolerance or not)
2. Collection interval can be changed for any class, to do this use nworks UI - select Enterprise Mgr node, you see all the Global Collection Settings. Uncheck 'Use Defaults' and you can edit multipler for each Class. Increase the multiplier for Host Disk Details. Note - you will get less data points delivered, but they will be averaged over the entire interval, so still accurate.

With above methods you should find you have complete control over how much data is stored in Ops Mgr
Any questions let me know!

Cheers
Alec

[maekee]

Hi Alec,

If i right click for example HOSTDISK.Collect.totalWriteLatency - Overrides - Override the Rule - For all objects of class: DISK

I only get to override parameter Enabled, nothing else.

Where do i find Enterprise Mgr node?
Do you mean Management Server tab, and then Collection Interval? dropdownlist? But isnt that Events?

/Maekee

[Alec King]

Hi maekee,

It sounds like you do not have 5.5 fully installed....

1. in 5.5, all performance collection rules have override possibilities of Tolerance, and MaxSampleSeparation (as well as Enabled)
2. in 5.5, the ui is renamed "nworks Enterprise Manager", and you will see Global Collection Settings and Interval Multipliers when you select the Enterprise Manager node (on Enterprise Manager tab)

Are you running the 5.5 RC build of MP for all components? It sounds like nworks UI and MP at least are still on 5.0.....
You have to upgrade all components to 5.5 - Management Center service, Management Center UI, Collector(s), and import 5.5 MP to SCOM.

Cheers,
Alec

[Alec King]

BTW - also wanted to say, the dropdown for Collection interval, and the multipliers in 5.5 - all this applies only to PERFORMANCE data.
Events come through essentially in real-time, within a few seconds, as they are received from virtual center.

Cheers
Alec

[maekee]

Hi Alec,
You are so correct, dont know where i got version 5.5 from.
We are using the version we got from you 29 April. Its 5.0.3.248.

Can i decrease the tables in the opsmgrDB without installing RC1 version of 5.5

I want to get down the size of these

/Maekee

[Alec King]

Hi maekee,

OK, I understand if you want to wait for full release of 5.5, before you install it. Anyway, it's coming very soon

In 5.0, the ways to decrease the data you are gathering are
1. disable some collection rules
2. increase collection interval globally
3. disable some metric classes

For (1), you could disable some rules - I would suggest perhaps VMHost-Disk reads and writes (since we anyway collect kbRead and kbWritten). Perhaps also disable VMHost-Disk deviceWriteLatency and deviceReadLatency (since we anyway collect totalWriteLatency and totalReadLatency).
Of course you lose some visibility of data doing that - but you still are collecting the most important counters. And all Monitors will still work anyway.

For (2), you already saw the Collection Interval drop-down in Management Center UI. You could increase that - but note, that in 5.0 this will affect ALL metrics. It affects performance collection (for graphs) and also Monitors. With this method you still get all counters, just delivered less often.

For (3) you can turn off a data class completely. For example, you could turn of VMHost-Disk class. Use the nworks Management Center UI to do this, just uncheck the class name. However you would then have no metrics at all for Host Disks....I would not recommend that, these are important metrics. If you really want to reduce collected data, you might also want to look at the VM-level metrics. Since there are so many more VMs than Hosts, they generate more data....perhaps disable VM-DISK and VM-NET classes. This can be OK, especially if you have Ops Mgr agents inside most VMs - -they will already be collecting Disk and Network I/O, so there might be duplication there.

Hope that helps - and watch out for the notice from Veeam that 5.5 MP has released! It has all the new features you need to control this performance collection.

Cheers,
Alec