@joshodgers has written some blog posts about jetstress you may want to check out.
Anything you can add here @joshodgers
Actually one of our customer is having a nutanix cluster of 2 8 series node and 1 6 series node.because they want the same setup like their physical infra available currently.The jetstress testing was going on fine with the customer .but they want to remove one hard disk physically and run the jetstress test.but on doing that the test got failed and when they tried to add the disk back its showing invalid disk configuration.and tests are failing.
A few things, firstly have you seen this post regarding Jetstress testing on tiered storage devices?
I would recommend you review and follow the recommendations here.
Second: Checkout this series showing Jetstress performance testing which shows vMotion (and soon) failure scenario testing videos of Jetstress on Nutanix.
Third: If you have followed our recommendations and have N+1 setup, meaning 1 less Exchange server than total nodes in the cluster, you should simply be able to shutdown a CVM while Jetstress is running on that node and the test should continue and pass with impact of roughly N-1. e.g.: If you get 2000 IOPS in a four node cluster, you should see 1500 if you took one node out. This is what I have pbserved during these failure scenarios, the larger the cluster, the lower the impact. Importaintly, its a predictable and graceful degradation, not a "fall off the cliff" style issue.
Let me know how you go, I'll aim to post the rest of the video in the Exchange performance testing series in the next few days.
I go through the links you mentioned.Actually in our case we are using one 6 series and two 8 series nodes.so is there any problem with this different models in one cluster?.customer has done jetstress test on one exchange vm on this cluster and the test was passing.and they want to remove one hard disk from the node and do the testing .so one hard disk not ssd was removed from one node and did the testing then the test got failed .even after adding the disk back and did the testing the test was getting failed.
only thing which was seen in the alert is nic errors on all the nodes .so we opened case with nutanix for nic errors and the case is going on.
Also i tried to access the curator page by http://cvm-ip:2009 .but not able to access that page ,may be that port need to be opened in the firwall of our network.But for accessing curator page ,any steps need to be done.
Thanks for your reply.
There is no issues mixing nodes, in fact in many cases we recommend it when scaling or designing solutions with large mailbox capacity.
What is the Jetstress configuration? How many databases, what size per database, how many vDisks per VM (one per DB and one per logs is our recommendation, with In-Line Compression for DBs and no compression/dedupe/EC-X for logs).
If you can provide this info to our support organisation, let them know to contact me and I can provide assistance and see what is happening.
If you are getting network alerts its possible a network bandwidth/configuration or teaming issue is causing the issue, but I can only speculate at this stage.
We'll get it resolved.
Thanks for your reply.Actually for the testing in your blog its mentioned compression should not be enabled.so we need to enable inline compression on DB container or not.
ok.josh.Also regarding the curator access page http:cvm-ip:2009.any command need to be executed for access this page or we can directly access this page.
Access to all the advanced monitoring pages including stargate are disabled by default, to access them you need to disable the firewall on CVM/s for each individual port/s, in this case 2009.
I personally wouldn't worry about viewing 2009 at least for this issue, we know how Exchange/Jetstress behaves, if the cluster/networking etc is working correctly, Jetstress will pass as per my blog posts.
Let's follow up in the support ticket as we like to track all issues so we can better assist customers in future support requests.