VMware Cloud Community
prashant_s
Contributor
Contributor

VSAN - PERC

Customers query is, can he have 1 server controller in RAID mode and other server controller in HBA mode in vsan and in same cluster.

Is this a valid setup?

What will be the implications if it is configured in this way.

Reply
0 Kudos
7 Replies
TheBobkin
Champion
Champion

Hello prashant_s,

I think the question should be more along the lines of 'why would someone want to do this?'.

Is there a mix of different controllers in the cluster or are you asking based on changing configuration(e.g. switching from RAID0 to Passthrough)? For the purpose of transitioning the clusters controllers from one to the other mode it shouldn't break anything. But you should be aiming for homogeneity of hardware and configuration of all the nodes in each cluster - any deviation from this will potentially lead to slowing down the whole cluster as they all wait for that one less performant node to catch up.

Bob

Reply
0 Kudos
prashant_s
Contributor
Contributor

Hi Bob, actually let me explain customers scenario here.

1. Customer is having 4 node vsan cluster.

2. Controller card personality is set to RAID.

3. He is facing some disk reset issues for which VMware suggested to change the controller personality to HBA as per VSAN HCL.

4. Customer wanted to know whether he can keep one server in HBA and rest 3 in PERC mode just to see whether the reset issue will reoccur.

Reply
0 Kudos
TheBobkin
Champion
Champion

Hello prashant_s,

What controller is in use?

If it is a H730 and in RAID0 mode then this is just asking for problems.

"Customer wanted to know whether he can keep one server in HBA and rest 3 in PERC mode just to see whether the reset issue will reoccur."

A better idea would be to get logs from before and after changing to passthrough mode and compare number of Power-on resets over an equal period of time (under same workload circumstances). You would generally note these in /var/log/vobd.log.

Bob

Reply
0 Kudos
prashant_s
Contributor
Contributor

The controller presently in the system is H730 and all HDDs are in non-raid mode.

Reply
0 Kudos
mklog
Contributor
Contributor

Hi Bob,

Our setup is having 4 nodes and all are having PERC H730 card and all HDD are in non-RAID mode and PERC controller personality is in RAID mode. Since, we are experiencing disk reset as per best practice we wanted to change PERC card personality from RAID mode to HBA mode, since our setup is in production we planned to work on one node at a time, hence need expert advice that if we proceed on the changes, whether there will any impact or is will work with mixed mode RAID & HBA for certain period of time, till we change all the nodes to HBA mode.

Reply
0 Kudos
TheBobkin
Champion
Champion

prashant_s this isn't a supported configuration so change this - if your customer wants to go through an exercise of second-guessing supports recommendations they can go and do that by themselves (outside of support).

I would advise also checking that you don't have other things misconfigured that can be factors in PoRs e.g.:

- Disks used for VMFS datastore on same H730 controller as the vSAN disks - unsupported.

- Disks being used for logging on same H730 controller as the vSAN disks - unsupported.

- Disks/partitions configured for coredump partition on same H730 controller as the vSAN disks - unsupported.

- VMFS space not removed following install of ESXi on same H730 controller as the vSAN disks - unsupported (though supported for *just* boot when VMFS removed properly).

https://kb.vmware.com/s/article/2136374

- Ensure using latest driver/firmware on vSAN HCL for these controllers, currently: lsi_mr3 7.703.18.00-1OEM.650/25.5.4.0006

- Ensure don't have the 1GB cache on the controller misconfigured e.g. using it for any % write etc.

Bob

Reply
0 Kudos
TheBobkin
Champion
Champion

Hello mklog

Check all of the points I outlined above in case there are other things you need to change also.

" since our setup is in production we planned to work on one node at a time, hence need expert advice that if we proceed on the changes, whether there will any impact or is will work with mixed mode RAID & HBA for certain period of time"

There shouldn't be any impact providing you do do it correctly e.g. make sure you have back-ups before doing anything, check all Objects compliant with their storage policies, extend clom repair delay timer to longer than 60 minutes if you think maintenance will take this long, use appropriate Maintenance Mode depending on your preference, allow to sync and check the state of your data between doing each node.

Bob

Reply
0 Kudos