Celerra File System Migration / Re-Architecture

Question

There are a host of design issues inherent to our current Celerra/CLARiiON environment.  Namely, the filesystems (primarily for VMware Datastores) are laid out terribly and need a great deal of shifting. We are planning to buy an additional shelf, fully populated with FC disks to hopefully remedy some of our I/O/performance issues which FLARE 30 should play a role in.  All of that being said, an estimated layout is as follows: clar_r5_performance: 2.7TB 98% used clar_r6: 2.1TB 98% used clar_r5_economy: 2.1TB 98% used clarata_archive: 8TB 80% used First_5: 1.8TB 30% used So, the plan is to create a New_Group, hopefully with approximately 20 FC disks.  After presenting to the Celerra, migrating all filesystems from the aforementioned FC pools to New_Groups, removing disks from the old system-defined, and adding to the newly defined New_Group for as large a spindle count as possible, utilizing FLARE 30. My question, 'in short', is how do we get there? We have a Replicator license so from what I've seen, nas_copy seems to be the most logical.  Is there any way to avoid a production outage?  Should any reconfiguration be necessary on the servers themselves with this method? Additionally, how would we go about extending the CLARiiON disk group without breaking the Celerra interface?

Rainer_EMC · Answer

extending Clariion raidgroups that are used by Celerra - DONT

you can use Replicator if the fs sizes stay the same and keep the downtime to a minimum - however since the file handles are going to change you would have to remount which means shutting down the VMs first

why not use Storage VMotion for your reorg?

Rainer

dynamox · Answer

you are going to 'pool' the new devices on Clariion or present regular flare LUNs  ?

tbaze · Answer

We will be Storage VMotioning where we're able but there are some CIFS/NFS/iSCSI FileSystems that are not VM-related and we need solutions for those as well. Is there any impact on Windows clusters re: disk signatures utilizing nas_copy?

We will create a new storage pool, copy all of the file systems to the new pool, and be left with .. approximately 40 disks. We would like to somehow combine all of these per the new ability granted by FLARE 30. I suspect this should be something EMC will need to support given that it's necessary to configure FAST/etc.?

dynamox · Answer

if you use Celerra Replicator to replicate iSCSI LUNs, disk signature will remain identical so MS Clusters will be happy.

Rainer_EMC · Answer

please make sure you get some guidance from EMC on what can and should be used with Celerra file systems for example using Clariion FAST or LUN compression is not supported for NAS LUNs

Rainer_EMC · Answer

I would suggest to use classic raidgroups and not Clariion pools when provisioning LUNs for Celerra use

Currently thick pool based LUNs aren’t supported for NAS use – only thin LUNs and only if the pool has only one tier and doesn’t use FAST or LUN compression

Rainer

Celerra

Was this post helpful?