Upgrade from 9.1 to 9.2 using ESX

[JGPSHNTAP]

Has anyone have any success doing an upgrade with a two node simulator in ESX?

We happen to be testing a D release of 9.2 but we are having issues with the network backend keep dropping out on us.

Any ideas?

[SeanHatfield]

Are you doing it from within ontap, or are you doing option 7?

[JGPSHNTAP]

Doing it from the CLI

[SeanHatfield]

I did a single node upgrade yesterday from the ontap cli.  'system node image get' took 5-6 minutes, but this is the phase where its downloading to mroot.  % complelte seemed to hang at several points, but mroot is on simulated disks so writes are expected to be slow.  Mine is on an SSD so it didn't take too long but on an HDD that could take a while.  'system node image update' went pretty quick.  With reboots the whole process was 15-20 minutes.

[JGPSHNTAP]

We did Single nodes with no issue.  We are testing 9.2D2

When we had two nodes together that is when all heck broke loose

[SeanHatfield]

IHAC that needs to do this as well, so I'll be trying that scenario shortly.  What errors are you getting?

[JGPSHNTAP]

the cluster database was out of qourum continuously.   Then once we thought we had it up, we lost one nodes back-end cluster network.    

We also had to increase vol0, which isn't a big deal.  We pushed 9.1 OVA"s 3x, clustered them in the config, and then managed to blow it up each time.

Hence, whey I need to keep it on single-node for testing

[SeanHatfield]

How much ram do your nodes have?

So far so good on my test case, but I'm not on the D patch.  How long did it take to fall over?

[JGPSHNTAP]

8GB of ram

sometimes it happened right away, sometimes it happen the next day

[SeanHatfield]

I'm curious now, so I'll put the debug patch on this pair and see if it misbehaves.

But, have you disabled snaps on the root aggrs and root vols? 

[JGPSHNTAP]

I don't recalling disabling snaps.. I will have to ask my co-worker.  Is that required?  

[SeanHatfield]

Simulator root aggrs are tiny, and a lot of stuff lands there now.  If it can't write to mroot it can't sync the RDBs.  Disabling those snaps can keep it from filling up when things like asup bundles, logs, stats, and config backups get created by built in scheduled processes.   You probably also have at least one image sitting in the mroot.  Check 'system node image package show' and 'df -h vol0'. 

[JGPSHNTAP]

I will let you know how I make out with the testing on single-node with the SIM. I'm specifically working with Development on this D patch.

[SeanHatfield]

For what it's worth, my two node on the D patch is still going.