What if it goes pear shaped ? These are things that used to keep us up at night. Driven by a requirement of quick and frequent updates, Dell EMC have improved the upgrade process for their mid-range considerably.
It is important to have a plan for your upgrades. Keep a regular cadence. I try and stay no more than 1 update behind, unless there is an urgent Out of band release or major feature release.
Dell EMC recently released version 4.5 for the Unity line. While not a major release, there were a few new features that made the upgrade compelling for me.
- Advanced DeDuplication
- MetroSYnc Improvements (including a control Application for site failover)
- File Level Retention (FLR) and a few more….
Beyond that, the release notes showed a significant amount of resolved issues. More so than I usually recall noticing.
I was also hoping that the VMware workflows (to automate the creation of storage objects) would create Datastores with VMFS6 format. I’m a little bit addicted to UNMAP !
So with a change control process approved, release notes read, I was ready to go.
The current OE version was 22.214.171.1244750794 and the Target version is 126.96.36.199.5.096
Open the settings and Software Upgrades.
Click on Start Upgrade and go ahead and run the Health Checks;
If there’s any issue, FIX THEM. You won’t be able to continue until they’re done.
Go ahead and choose the Upgrade package downloaded earlier and wait while it’s transferred to the array verified.
Decision time. If you’re certain you have your multipathing sorted, leave the “Automatically Reboot Storage Processors” checked. This will ensure that the SP’s are rebooted gracefully, and any storage resources are first migrated to the peer SP, avoiding any data being offline.
Notably for best performance in VMware ESXi, you should have Round Robin selected, with an IOPS limit of 1. See here.
Go ahead and click NEXT….
Review the summary. Note the warning about I/O Utilization. It’s good practice to ensure that your overall CPU utilization is below 50%
Why ? Because as noted earlier, each Storage Processor (SP) will be rebooted and for a short while, each remaining SP will be hosting all your storage objects. If the sum use hits 100%, you may observe some latency spikes.
When you’re ready, click “Finish”. Kind of odd, it should really be “Start” !
..and we’re off. Note the predicted time to complete is 92 minutes.
The first reboot is coming up. You had your multipathing sorted, right ?
If you did, you’d see “Path redundancy” warnings in ESXi, but the datastores would remain up and happy, like mine;
Also be prepared for a small barrage of warning emails;
We’re nearly there. Hang on tight.
..and done. Reload Unisphere. (F5).
Voila. All done.
Now, I suggest you review any errors/warnings. I also recommend to do another set of health checks and once you’re sorted, re-enable all your replication session that were paused at the beginning.
Now, I started this upgrade at 9.21am on a Monday morning and it was completed at 10.17am, just under 1 hour, and well under the 92 minutes initially predicted.
What’s the diff ? New “Pre-allocated Space” in the pool widget; and “Advanced Deduplication” in the object properties. Very nice.
…and the pièce de résistance “VMFS6 available in the workflows”
So, my advice? Go for it. Plan carefully and do your diligence.
These upgrades are smooth sailing and there’s really no excuse to not keep things relatively up to date.
Go at it.