RAID 5 and hotplug

[trib]

Hello,

I have a ML370 G3 with 3 disks, which was preconfigured with RAID5. I'd like to know what would happen if one of the disks fails, since there is no spare disk ? Would the system go on, but slowly until the disk is replaced, or would it simply stop working, waiting for a disk to rebuild data ?

Another point, about hotplug : I have another ML370 configured with RAID0+1, with an additionnal spare drive. I can hotplug the spare drive without any problem, but i'd like to know if I could do the same thing with one of the mirrored disks, while they are working ?
(This is for test purpose, i'd like to check myself what happens when one of the disks fails before moving the machine to production, but i'm not sure if the raid controler should have marked the disk as deficient before hotplugging it.)

Thank you

[JohnWRuffo]

Yes. You can "pull" the drives while the array is up. That is true for both RAID 1 and RAID 5.
The arrays will enter what is called "Degraded mode". While the array is in this state, it will run, but it will have to generate parity info on the controller for every read and write that happens. This will signifigantly slow the system down. Also, while in degraded mode, if you loose another drive, your array WILL fail. Be sure that you do not "remove the failed drive from the system untill you have a replacement in hand.

Also, do not remove/replace more than ONE drive at a time on any array. Let one rebuild and then another. Let the rebuild complete fully before starting another. If you do, there can be problems and more importantly, the array controller can only rebuild one at a time anyway.

I hope this helps you. Let us know if you need more info.

-john

[JohnWRuffo]

Yes. You can "pull" the drives while the array is up. That is true for both RAID 1 and RAID 5.
The arrays will enter what is called "Degraded mode". While the array is in this state, it will run, but it will have to generate parity info on the controller for every read and write that happens. This will signifigantly slow the system down. Also, while in degraded mode, if you loose another drive, your array WILL fail. Be sure that you do not "remove the failed drive from the system until you have a replacement in hand.

Also, do not remove/replace more than ONE drive at a time on any array. Let one rebuild and then another. Let the rebuild complete fully before starting another. If you do, there can be problems and more importantly, the array controller can only rebuild one at a time anyway.

I hope this helps you. Let us know if you need more info.

-john

[Doug Wogan]

I wouldn't recommend pulling any of the drives while the system is up unless your drives are dead. You can cause disk problems and possible data corruption when pulling an on-line drive. Hot plug drives are for hot "plug" not hot "un-plug". If you want to test power your system down, pull one of the drives and power back on. Once you're up and running it's safe to plug your hot-plug drive back in.

[JohnWRuffo · Accepted solution]

Doug is correct in that if you pull an "Active" drive, there can be data loss or the system can hang:

The the attached JPG...

[trib]

Ok, thanks for the fast responses.
So if i understand correctly in RAID 5 i can loose one disk and replace it say one day later with no stop but a big performance hit. Very risky since if one of the other disks fails data loss will occur.
Now, how many time does it take for a spare drive to be rebuilt ? let's say if configured at the fastest rebuild speed ?

Would the rebuilt be faster on RAID1 ?
As you understand my purpose is to have the maximum of data security, and to reduce the laps of time the array is not redundant. I don't mind about losing space, so i'd change for RAID1 if it's worth it.
But the problem is that this machine has been put in production without thinking about that, so it'd had to be really worth it ;)

Also about the hotplug: ok i can remove a drive while the machine is off and it will still boot, but it requires a physical intervention (to press F2). This would not occur if the disk had failed while the machine is running right ?
I suppose it should start reconstructing the mirror on the spare disk itself and drop me a line in the log.

[Sean Marshall_1]

The average speed of the Raid rebuild is 15-20 mins per gig of the array. This is if there is a BBWC (Battery Backed Write Cache) on the array controller. Without that it would be slower. Since Raid 1 (in general) has fewer drives then Raid 5, in theory it would rebuild faster.

Check out the page Page 3 - 7 of the manual at the following link http://wwss1pro.compaq.com/support/reference_library/viewdocument.asp?countrycode=1000&prodid=117&source=239449-004.xml&dt=264&docid=14346


Hope this helps.

[Grace_8]

I have a similar question.
I have a ml370 with 3 hotswap drives and have purchased an additional 2 drives to increase the storage space on this server.

Can I just slip them in when the server is down.. or do I need to change something in the configuation?

Thanks

[Sean Marshall_1]

You would need to add the drives, while the server is powered down. Then go using either the F8 on post or the Array Configuration utility, you would have to expand/extend the array/logical drive. However there are some things you should know about it first. You might want to check out this site first.

http://h20000.www2.hp.com/bizsupport/TechSupport/Document.jsp?objectID=PSD_EM000915_CW01


You need to update the driver of the array controller, the firmware of the controller (possibly the drives as well) then update the ACU that your using. MAKE SURE YOU BACKUP YOUR DATA BEFORE TRYING TO EXPAND / EXTEND. Hope this helps

[Grace_8]

Thank you.. I have read that information and I am not sure I understand something.
It says that there are 3 things that must be done in order to extend and I need some help with the following .....

"The Smart Array Controller must be flashed to the latest version of the Option ROM."
How do I do that?

(The other 2 things are sp3 for win2k and the new configuration utility.)

Any help would be appreciated .
Thanks