NFS connectivity issues come with ESXi 5.5 U1
Frequent NFS APDs after upgrading ESXi to 5.5 U1
So after catching up on some weekend tech notes, KBs & emails I came across this new admission from VMware Support about NFS connectivity issues from ESXi 5.5 Update1. This was very interesting to me as my firm is an all NFS storage shop and we are looking to move up to 5.5 from 5.0x during this quarter.
APDs to NFS volumes
This KB from VMware discusses the APD (All Paths Down) states that are experienced by those running VMware’s newest ESXi update, 5.5 U1 which was recently released the week of March 11th, 2014. This issue also affects those using VSA Datastores as the VMware VSA uses NFS and not iSCSI. As I read through the document I thought to my self, ‘awesome…just what I want, more issues to deal with at work and right after I planned to do this beloved upgrade from 5.x to 5.5!‘
As I read further down it definitely seems that this issue according to VMware is not part of ESXi 5.5 GA as their current recommendation is to use this version to avoid the issue.
Quoted from KB 2076392:
"This is a known issue affecting ESXi 5.5 Update 1 hosts with connected NFS storage. VMware is working towards providing a resolution to customers.
To work around this issue, VMware recommends using ESXi 5.5 GA."
Part of me is still not 100% confident though. Call me crazy but I have experienced some odd NFS disconnects with my ZFS based NAS where either rebooting the NAS, the ESXi Host or both were the only way to get resolve. I know VMware has corrected some NFS issues in 5.1 U1 relating to VSAs in Maintenance Mode but I still feel NFS gets treated like the red headed step-child when compared to it’s block level big brother, iSCSI. Oh well I guess me and my NFS shares will have to take that back seat at times.
What version and when?
Next I began to play out my upgrade path to a version higher than 5.x. For me, my VMware View (or Horizon View) is the most important piece of our infrastructure to consider 1st. We currently run version 5.0 of View, the version before it changed names to Horizon. We would like to get the hosts up to 5.5 and View running 5.2 or 5.3 depending on if we pick up more pieces to the Horizon Suite or not. After reading this KB on NFS issues in 5.5 I thought about just waiting it out even longer. But by then I might see Horizon 7 pass me up! LOL.
Final Thoughts
I think the best plan would be to stick to my current thought, move hosts to 5.5 (skipping 5.5 U1 to avoid any new NFS headaches), View to 5.2 at least and then patch up the 5.5 hosts with that cool new OpenSSL fix. Yeah, Heartbleed, have to worry about that too now on 5.5. Please don’t forget to patch that up. Yes, Yes we all know that isolating your Mgmt Network will possibly cut those SSL bugs out of the picture, but just do the patching anyway. Safe is better than Sorry. Also note the details on the OpenSSL bug fix that can found here. Patching 5.5 for Heartbleed and then upgrading to 5.5U1 will revert your systems to a vulnerable state, DOH!
Note: After you have patched your ESXi hosts with VMware ESXi 5.5, Patch Release ESXi550-201404020, you should not upgrade your hosts to ESXi 5.5 Update 1 as the hosts will again we vulnerable to the OpenSSL Heartbleed issue. After applying VMware ESXi 5.5, Patch Release ESXi550-201404020 on ESXi 5.5 hosts, you should only patch your systems with VMware ESXi 5.5, Patch Release ESXi550-201404001 to update your hosts with all bug fixes that were provided with ESXi 5.5 Update 1. For more information about this patch release, see KB 2076586.
Here are some links to follow along. Thanks for checking me out!
Frequent NFS APDs after upgrading ESXi to 5.5 U1
Resolving OpenSSL Heartbleed for ESXi 5.5
Update sequence for vSphere 5.5 and its compatible VMware products
Post image borrowed from: http://cormachogan.com/2012/11/26/nfs-best-practices-part-1-networking/