vSAN/vSphere 6.0U1 to 6.5 U1 Migration Best Practices ??

[rking4067]

Hi,

For a number of reasons we are using vsan 6.1/vsphere 6.0 U1 on several 3 node clusters of Dell R630s.  We are hitting a number of latency

issues with a relativity light workload.  Seems like a logical time to get upgraded to more recent releases given all the improvements to vSAN.

We know we need to also watch HCM and dell firmware updates are a must.

Can anyone share best practices or things to watch out for with such an upgrade? 

Thanks in advance.

[TheBobkin]

Hello rking4067,

Are you seeing a lot of H:0x7 (RESET) and/or H:0x8 (ABORT) Sense codes or 'Power-on Reset' in the vmkernel.log?

I ask as I am assuming these are running on some variant of H730 controllers and the older drivers/firmware on these had a plethora of issues.

Note that these controllers have non-defualt (default in 6.2) configurations that can assist with some of these issues (and don't require reboot):

kb.vmware.com/kb/2144936

Also note that having VMFS on the same controller as vSAN disks is unsupported and problematic on H730 with lsi_mr3:

kb.vmware.com/kb/2136374

As 'Full Data Migration' Maintenance Mode option is not possible in a 3-node cluster, I would strongly advise that you take good back-ups of everything before proceeding with this.

Ensure that all Objects are healthy and compliant with their Storage Policies.

"Can anyone share best practices or things to watch out for with such an upgrade?"

After upgrade hosts can sometimes take significantly longer to boot with longer periods of 'SSD Initialization', regardless of how long this process may take it is not 'stuck' and should not be rebooted.

Multiple on-disk upgrade will have to be performed on each of the hosts (to which version depends what on-disk version you want to upgrade to) this will require rolling removal of disk-groups and recreation on the higher versions, whether you want to do this later or at the same time is your call - but again make sure your back-ups are good as it will be running on  reduced-redundancy (e.g. only 1 of 2 data-component replicas available).

Bob

[rking4067]

Thank you.  Yes we know we are running old firmware as well on these systems.  As I am sure you can appreciate many of our challenges are logistics, testing and outage window related.  We have other reasons to upgrade besides just latency.  This is really good information, thank you!! Very helpful!