This post is more than 5 years old
2 Posts
0
31228
Replacing all disks on ps4000
Hi everyone,
we have a bunch of ps4000 with unsupported disks , which we want to replace.
we have 8 slots in use , and therefore 8 slots free.
can i add the replacement disks in the 8 free slots , create a new storage pool and migrate the member to it? Would that move all the data to the new storage pool including volumes without downtime for our vmfs volumes?
or would i need to create new volumes in the new storage pool and just do a storage vmotion?
is it possible at all to have a second storage pool with different size disks and maybe different raid level?
look forward for any feedback.
thanks,
Sergio
DELL-Joe S
7 Technologist
7 Technologist
•
729 Posts
1
December 20th, 2012 07:00
If you have a support contract for your arrays, you may be able to contact you Dell sales rep to see if they can get you a loaner array.
Another option would be to see if you have a group that has very little data on it (and/or low usage). Take the data off of this group (backup/restore to another group). Or instead of the backup/restore, you could create a volume in another group, and attach it to the same server via iSCCI, then move the data that way (for example).
Once you have the data off of one group, you can reset that array (has no data now), reconfigure it with the new disks, and then use it as your “rolling” update array (i.e., add it to the group and same pool for the arrays you need to update the disks on). This method will take some time to compete; add the member, wait for raid rebuild and data balancing, move the other member to temp pool, reset the now empty member in the temp pool, add disks, reconfigure, add back into the group, etc.).
BTW, since these members have unsupported disks, I would prefer that you reset the member once it’s in the temp pool (delete member from group), as opposed to replacing each disk one at a time. I’m not sure how the firmware will react with unsupported disks, even though no data will be on them once moved to the temp pool.
-joe
Anonymous
5 Practitioner
5 Practitioner
•
274.2K Posts
1
December 20th, 2012 11:00
Forcing the rebuild to replace the disks, isn't a good idea. Since you are forcing a rebuild, the possibility of ending up with a faulted array, should another disk fail, is much greater. Especially with un-certified drives. If you do decide to do that, make sure you have a validated backup first.
DELL-Joe S
7 Technologist
7 Technologist
•
729 Posts
0
December 20th, 2012 06:00
Do you have more than one array member in the same group? If so, are the members in that group in the same pool? If so, is there enough free space in the pool to hold the data of the member you need to replace the disks on?
First, you cannot “split” an array into two pools, an array can only be part of “one” pool at a time.
Each array can only have one raid policy, so adding the additional disks into an existing array (with only 8 disks) will expand the existing raid policy of that member to the new disks. You would end up with a 16 disk array (where all the volumes on that member will span across all disks).
Basically, the process is (if you want to continue to only keep 8 disks in each member); you would need to have at least two members in the same pool, and enough free space to remove one of the members from the existing pool to a temporary pool, once in the temp pool, you can then remove the member from the group, and then perform the disk swap.
When moving the member to a temporary pool, the volumes hosted on it will remain in the original pool and you will have no down time. Once its in the temporary pool, you can either vacate the member and reconfigure it with the new disks (fastest, but you would have to setup the entire array scratch, like the eht IP address's, etc.), or you can systematically replace each disks (remove, insert, wait for the rebuild to complete, move on to the next disk). The second method is much slower, and will take a few days to complete per member. Once the disks are swapped, you would move the member back into the original pool wait for the load balancing to complete (could take another 24-72 hours (or longer depending on the amount of data, load on the group, etc.), and then repeat the process for the next array.
I would suggest that you give support a call so they can help you plan this all out.
-joe
sergiof
2 Posts
0
December 20th, 2012 07:00
its one array per group.
pity it does not work the way i hoped.
so the bottom line is to get a temporary unit to do the swap :emotion-1:
thanks for your help.