Monitor DISK IO latency

[varrus999]

Veeam Monitor 4.0.

We are monitoring a number of data stores on shared storage (SAN EVA 4400). I can monitor disk space, and have tested the alerts ,but I am trying to monitor for disk errors and disk IO, particularly latency. I don't see any data when I look at reports for disk IO or disk issues. Can someone tell me how I can verify that this is working? Right now it says 'no data available'

Mike.

[Gostev]

Mike, please upgrade to the latest version and see if this resolves the "No data available" issue.

[varrus999]

IS this a simple upgrade - will it keep the existing database and rules intact?

Thanks for the response

Mike.

[varrus999]

I read the release notes to upgrade and now the collector servcie won't start and I see nothing in the console. Called support...now I'm waiting for a tech to call me back..

Has anyone else had this issue?

[varrus999]

Never mind...found this post and did the same thing - VEEAMDCS was 4.0...found the 4.5 version with the .ex_ extension in the install folder. It took a long time during the install to 'stop the VEEAM Data Collection Service"...that should have been a clue that it wasn't able to do it - should be an error in the install at that point. Oh well.

Re: Veeam Monitor 4.5 Released
by PDMeat » Mon Oct 12, 2009 1:43 pm

I had a problem on install that I eventually figured out was simply that the Veeam Data Collector service exe file never got swapped. I stopped the VeeamDCS, manually renamed its exe and grabbed the new one from the 4.5 install, restarted the service and it works well now.

[varrus999]

Am happy to say I now see disk IO data starting from when I upgraded (actually I think it grabbed the last hour's worth from the ESX hosts).

A few questions
-this data will be dumped into the database for historical reporting, right?
-Obviously the data needs to be interpreted. I know we can look at trends to see of we are getting spikes, increased IO activity once we have some data and establish a baseline, but where can I find a way to correlate this data against real world values - IE. what would be considered good and bad when it comes to monitoring something like DISK WRITE latency? Where should we set alarms - how many milliseconds for warning / error, etc. I know this is VMware knowledge..am wondering if there is any best practice guides out there?

Thanks again,

Mike.

[Gostev]

Hi Mike, I am glad all working well for you now.

Yes, this data is stored in the database for historical reporting.

Recommended warning level is 60ms, and error level is 100 ms for disk latency alert. We are currently working on building this knowledge into the product, so hopefully with the next release you will have much more pre-defined alerts with the recommended thresholds, as well as knowledge base available directly from our product.

Thank you!

[varrus999]

Thanks again. Do you have any documentation / whitepapers on this?

Mike.

[Gostev]

Not yet, but we are working on this for our next release.

[varrus999]

Sorry to clarify one point - recommended settings of 60ms / 100 ms applies to write, read, or both?

Mike.

[Gostev]

Mike, its both.

[varrus999]

Thanks for the quick replies. It looks like you are busy monitoring the forums!

Mike.

[schwagro]

I am trying to get these numbers on a per VM basis. The problem is that when I choose Disk Latency (Read or Write) i see the numbers per host. When i edit the counter to display per VM, it shows 'No Data Available'.

Also, when I vew them at a Datastore level (per LUN), my LUN names are very long and ugly. Any way to get them to display the neat label for them that is in vSphere?

[Vitaliy S.]

Hello,

As far as I know Disk Latency counter is only available for host objects at the vCenter DB, so it is not possible to choose per VM basis. Veeam Monitor displays all datastore names as you see them using vSphere Client. So could you please clarify the LUN name issue with the screenshot?

Thank you!

[schwagro]

Sure, here's where I'm seeing the issue.

[Vitaliy S.]

Hm.. actually all these names we recieve directly from VMware through API calls, but I need to consult with our dev team on what we can do to help you here. I will update the thread as soon as I have any information

[Vitaliy S.]

Well... on the right picture we show you all devices that relate to DEV and PROD objects and those names are the same as shown in vSphere. And if you click on any object on the right picture it will take you to the parent object the device relates to. But thank you for your feedback!

[schwagro]

Even if you do that though, it shows server names, but the disks they are attached to still have those strange names. I would prefer to have everything in one shot with readable names. Just something to pass on if it's possible to do.

[afidel]

I just wanted to chime in that I too would find it very convenient if the naa-* addresses could be mapped to their coresponding friendly names in the detailed views. I have a number of servers with disks on a number of different datastores due to performance tiers and would like to be able to quickly figure out which disk is which.

[Vitaliy S.]

Thanks for the input! Starting from Monitor v5 you will see user-friendly names of your devices along with their IDs that might be needed for troubleshooting.