We got some new disks for our CX3. Instead of creating a new RaidGroup with these extra disks, I want to add them to an existing RaidGroup. In other words, I want to Expand that RaidGroup.
Never done this before. I expect it to be similar to expanding an Array on a server. It will probably run a rebuild, which will take some time considering the size.
The LUN's on the expanded RaidGroup don't have to be enlarged. I just want to use the extra space and create additional LUN's. So no MetaLUN's.
So what are the risks ? Is it perhaps better to empty-out the RaidGroup (move the data to another RaidGroup), so it can be wiped and then re-created including the new disks ? That's a lot of work though.
Also, there's a related issue.
The RaidGroup I want to expand also has to be defragmented. It seems a LUN has been deleted leaving a 'hole' which doesn't allow all the free space to be used, only the biggest continuous space.
Do I have to run the Defrag seperately or will this be done automatically during the Expand operation ?
Any suggestions are welcome.
A few questions you should think about (help you to understand any risks) :
Because the LUNs are being re-jigged, and I guess sometimes SH** happens, I've never had a problem but always good to make sure backups work and are upto date.
Performance impact to the applications and the Clariion when you run the expansion and the defrag - depends on what the Clariions doing now, how many users, apps etc.
Long term risks
I would have a look at the following white papers:
1)Clariion best practice Flare 28.5 (what ever Flare code you have, so there is a 26 document as well)
3)Clariion Hard Drive technology
Could your expansion effect the performance of the Clariion or the application(s) in an adverse way. Have a look at the questions below, and have a read through the above papers for best practice guidelines.
There are good and bad ways of architecting an array. Adding more disks to an existing RAID group, and creating more LUNs in there may or may not be a good thing to do, and those papers should give you a good idea on best practice, design.
for example, you could end up with too many spindles in your RAID group, which would effect the rebuid times, if you have a disk failure.
questons to ask:
1)what RAID type is the original Raid group (e.g.RAID 5, RAID 1/0) ?
2)How many spindles are in that Raid Group ?
3)what type of disks are in the original RAID group (e.g. type Fibre Channel, 15k).
4)What are the new disk types - same as the above ?
5)How many disks you adding in ?
6)How many LUNs are there at the moment, in the RAID group
7)How big are the LUNs (GB), so how much data we going to be re-distributing ?
8)What kinds of applications are using those LUNS, any databases, any virtual machines, or just files etc.
Expanding a RAID Group does not automatically increase the user capacity of the LUNs, it distributes the capacity of the LUNs equally across all the disks in the Group, freeing space for additional LUNs.
I can't rememeber whether the Clariion would de-fragment as well at the same time, but I don't think so. I I think you will have to additionally defragment as well, after expansion.
Have a look at the Navisphere Manager Administrator Guide in Powerlink (pdf), it's a nice reference manual.
anyway, answer the questions I've listed, and I'll have a look as well for things you might want to watch out for.
What flare code are you running? I'd assume it's somewhat recent (>25?)
I've never had an issue doing what you laid out.. adding disks to the raid group is pretty easy process and goes fast. The "transitioning" of the LUNs belonging to that Raid group will take awhile though, but everything can be used (online) during that process (you'll see the big "T" on the Raid Group and LUNs during the transitioning).
You will have to do the defrag seperately.. adding disks to the Raid group does not fix the "hole" that's already there, but it does give you additional contiguous space (because you're adding more space with the additional drives).
Since the above is still two seperate processes, I'd probably do the defrag first and then add the disks to the raid group.. in my mind that'd seem more streamlined from the CX's point of view.. either way though, there will be quite a bit of time "transitioning"..
We're running Flare 3.26.
The RaidGroup currently consists of 5 SATAII spindles of 1TB in Raid5, so adding 4 more disks (same type) will create quite a big RaidGroup of 9TB raw space. Perhaps this is too risky by itself considering the amount of time a rebuild of this will take. It made me think twice anyway.
The LUN's (15 in total now) on that Raidgroup are anything between 100-500GB and are used for VMWare VMFS and Windows SAN attached LUN's. Applications on that RG are mostly low I/O, since SATAII is not as fast as Fibrechannel. So file-services, archive, backup and low I/O VM's in ESX.
I might skip the expansing and create a new RG anyway. There's not that much time left and we need the new space.
Not so good news.
Before I went ahead with the expand I decided I wanted to try it out on a non production RG, since I had 4 empty disks to play with.
So I put 3 disks into a new RG using Raid5 and created 3 LUNs on it. I then added the 4th disk by running the expand. Indeed the expand went rather fast but the Lun transition not. In fact, it's still running as we speak. Yep, on a 4 TB RAW raid-group it's been running for 24 hours now. The expand progress indicator even says it's still at 0%. Priority is at 100%. This can't be good.
Glad I didn't do this on a production RG, which I honestly have serious doubts about now.
I removed all the LUN's and destroyed the test RG. I'll try the same trick one more time with the unused disks.
Have a look at emc134339 Does a RAID group expansion defragment the RAID group on a CLARiiON storage system? in the EMC Knowledgebase (IView from within Powerlink). According to this, a RAID group expansion will also defragment the RAID group. It will happen as part of the RAID expansion process.
(I learnt something new today!).
I do think in your case it would be better to just build a new Raid 5 (4+1) with the new disks, and then create the new LUNs in there, you've already alot of LUNs in the existing RAID group. If you decide not to expand the existing RAID group, you will need to run a defragment on it so you have all the unallocated space available.
Do you have any virtual machines in the old RAID group (you've some VMFS volumes configured) ?
You said you're also using the old RAID group for some archiving and backups, are you using RAW LUNs here ?
It's worth while reading The white paper 'EMC Clariion Integration with ESX Server'. It covers Raid Group and LUN design for VMware environments.
Time to expand
Just curious if you tried you're test 'expand' again and how long it took ?
I've had a look at some numbers/speeds - I found lots on Disk rebuilds (for different types of disks), eg a 500GB SATAII drive rebuilds in about 4.5 hours, thats with an array that has no workload, on a setting of 'High'. Haven't found much on speed/time taken to expand RAID groups /restripe the data.