OneFS 7.2.1

I am kind of in the same scenario. I need to upgrade because of new S210/NL410 nodes. I am currently running 7.2.0.2, which has worked almost flawlessly. We are also running mixed protocol (FTP, NFS, a little SMB).

Is the release stable or going to be marked as Target Code any time soon?

The guidance I have been given on the next update to it (7.2.1.1) is in a few weeks.

7.2.1.0 has significant issues. We haven't gotten to the bottom of them yet.

There seem to be major performance issues if map.lookup.uid is turned on cluster wide.

The celogs process is extraordinarily slow and the mount error throttling doesn't respect changes made via gconfig.

The network driver issue is horrifying (either you get nodes rebooting randomly, or you get data corruption. Your choice!) (see patch-162062, ETA 208773)

It sounds a bit like you have been a bit unlucky in your environment. Sounds good 7.2.1.1 is coming up soon! Have you got a changelog?

We are not using map.lookup.uid although we are running NFS, we tend to stick with "normal" linux practices regarding permissions. We are not running with windows permissions at all on the cluster as we are "more or less" entirely using unix based systems. Protocols are NFS, FTP, SMB. I have never found shared permissions necessary or very well implemented. I did test it at some point, and the results using Isilon were much better than IBM systems we tested in the past. A lot of work seems to have been done to make this work well in OneFS, but you knever know in a mixed environment. So I tend to stick with my linux permissions. They may not be as flexible, but they always work as expected.

Regarding "patch-162062", this is an well know problem that's been around for some time (patch shows 7.1.1.3). Luckily for us, we are not running MTU higher than 1500 currently, as we have some network devices not supporting this all the way down the tree. Therefore we should not be impacted by this issue, but thank you for pointing this out. I bet this had you banging your head, once or twice.

I think I will give the update a change, and let you know my finding.

Nope, no changelog.

We don't use windows permissions on our cluster either. So much easier to administer. map.lookup.uid actually just moves the group lookups from the nfs client to the Isilon cluster, so it eliminates the 16 group limit and provides marginally better security.

We were actually not affected by 162062, as we finally were told (after I had installed it and done the completely unnecessary cluster-wide reboot) that the data inconsistency is only triggered when tagged vlans are used. I reverted the driver to the earlier version by hand with kdunload/kdload, avoiding a second reboot.

How many node types and what quantity of nodes do you have?