Hi all, hopefully someone can assist with this query regarding the HCIBench tool.
We have several VSAN clusters (vSphere 5.5/VSAN 5.5), one of which has been pencilled in to be a dedicated VDI cluster. Before we deploy to it though we want to get some stats out of it to determine where the stress points are likely to be and what the best VSAN stripe width will be (etc).
So, cluster is configured with a VSAN datastore which is up and running, tested and functioning fine. However, we're having trouble with the HCIBENCH tool when performing a validation. It simply comes back each time and says that the clusters isn't VSAN enabled (which it definitely is). Thinking this was just a minor bug I've clicked the TEST button anyway and the pop up window comes up for about 5 seconds then returns 'test completed', no VMs have been deployed and several empty result folders. However, a few minutes later the VDbench VMs do start to get deployed in the cluster and powered up.
Is this normal behaviour for HCIBench or is there something I should be doing differently, I don't see what else there is to change particularly, all seems pretty basic to configure. Just doesn't seem to execute (& view results) cleanly.
Any help appreciated.
I'm actually in the process of creating a step by step post on getting HCIbench to run, as I've had a few hiccups myself.
Check out the following:
1. Log onto the appliance, and check the config doesn't have any funky characters - you'll find it in opt/automation/conf and is called perf-conf.yaml.
2. Check out the logs in /opt/automation/logs and see what it reports when you try to do the validate
Let me know what you find. I can include it in my post if you think others might hit it.
Thanks for the quick response Cormac, I've not found anything odd in the conf file and the logs aren't showing anything exciting either.
Having said that, viewing the file system in the appliance has helped me to understand how it hangs together a little and I suspect the errors and odd behaviour I'm seeing are just related to the web form (or user error). For example, I don't see anything in the drop down menu for 'Select a VDBENCH parameter file', not even the 'Use All'. But I do see them in the file system when I log in to the appliance and when I actually run a test, all the VMs are getting deployed, the tests are run, results are saved to a zip file and the VMs cleaned up after. The results zip file also contains a sub-folder for each of the vdbench parameter files that I can see listed in the Param path on the appliance.
So, it does appear to be doing the important stuff of creating the VMs, generating the load and clearing up after itself. I'll keep checking the logs and if anything pops up I'll post it on here.
i assume the vdbench zip file was downloaded from Oracle and uploaded to the appliance?
when I did that step, I then went ahead and built my own parameter file.
Hi Cormac, yeah the vdbench zip file is uploaded. Building the parameter file seems to work and it shows it in the appliance when I log on. Just doesn't list it through the Tool Configuration Page when I drop down the menu.
Hi Cormac, yes I clicked the refresh button a few times. Each time it reported 'Refresh OK' but still nothing in the drop down... However, when I opened the config page this morning the drop down list now has all my configured tests listed (although the validation still fails). The tests I had running last night all completed too. Not sure what the issue is but at least I am getting test results out.
cluey , CHogan ncriverside : I have a hybrid array I am trying to run One of the volume as a VSAN data store . Is it required for me to check if they (their controller) support pass through ?
I am suspecting that to be the reason why I am seeing my array LUNs as ineligible disks ?
hi Oopszy, I'm not 100% clear on your environment and what you are trying to achieve, if it's what I suspect I don't think it will work too well.
with regard to a storage controller and pashrough, pass through mode isn't essential for the controller but it is preferred. If it doesn't support pass through/jbod then you can use a RAID0 mode, but this has a few limitations. You have to inform VSAN which disks are
your cache (SSD) as VSAN will not have direct access to the physical disks and cannot see which is an SSD because of the RAID layer sat above the physical disk. I've never done this but as I understand it you create several RAID0 volumes each containing just 1 disk. You then inform VSAN which of these single disk RAID0 volumes is the caching disk from the command line.
Best bet (especially in production environments) would be to check the VSAN HCL and make sure the kit is listed. I'd stick to pass through mode if possible.