Define best 

Easiest and safest way imho is to NOT build a new cluster... but join the new nodes and let OneFS do all the work..

Will probably need to swap in the new IB switches and update OneFS before joining the nodes.

For commercial assistance, look out for Superna Eyeglass as listed here:

SyncIQ - Isilon Info Hub

And a free tool which can be helpful:

(add http) github.com/obergt/Isilon_Tools

hth

-- Peter

Best way means safest way.

A downtime is not possible because the (old) cluster is in use and cannot be switches off or restart.

Because the cluster is running OneFS 6.5.x a rolling update to 7.x.x is not possible, correct? That means a simultaneous upgrade with a little downtime and disconnected clients would appear.

The actual plan is to build up the new cluster parallel to the old one and create (or import) all SMB-Shares, local users and SmartQuota settings from the old to the new cluster. The next step is to copy all data from the old to the new cluster directory for directory. Next the active users will be switched bit by bit from the old to the new cluster by changing the mounts of the clients to the new cluster. Every department can switched successively from the old to the new cluster.

I see. Difficult call... might or might not be worth the hassle.

If you are going the clean up some "grown" idiosyncrasies,

this will speak in favor of the step-wise procedure...

On the other hand, you must be Masters of the DNS

to migrate one department after the other.

Good luck (and let us know the outcome : )

-- Peter

SmartQuotas can be exported to an XML files and imported on the new one. If path on the new cluster changes you can modify XML to reflect that before import. We had 1500 home directories with directory quotas set on them to be migrated from 6.5 to 7.1

Phillip,

If you haven't seen it already, you might have a look at Superna Eyeglass for Isilon.  Your EMC Account team can engage them for you.  6.5 to 7.x is not their normal use case, however I know that they have been looking at it.  Ultimately it's meant as a DR solution, but it's relative low-cost and flexibility could certainly be helpful to use in your implementation as well, and then perhaps be continually used to replicate this type of configuration information to a DR cluster.

The alternative if your old hardware is not really that old, is to not use SyncIQ to replace the cluster, instead upgrade the old cluster in-place, then add new nodes, and smart-fail out the old ones.  This causes a much shorter cutover window, and you don't have to re-configure anything:

1. A simultaneous reboot for the upgrades

2. Moving over the client network interfaces (this impact can be heavily mitigated with proper planning)

This could give you a much better overall experience.

Superna - Eyeglass Isilon Edition

~Chris Klosterman

Senior Solution Architect

EMC Isilon Offer & Enablement Team

chris.klosterman@emc.com

Thank you to all,

I've discussed the possible solutions with the customer but he don't want to and cannot go these ways.

Is there no other way?

Is it possible to downgrade the new cluster to the same OneFS version 6.5.x, than export/copy the settings (e.g. /etc/passwd for the local users) to the new cluster? If you have two same version clusters there must be a way to use some of the old settings.