Urgent Help - Linux - HP EVA 5000

[Sandy_42]

I have HP EVA 5000, Windows 2000 Clusters & Red hat ES 3.0 in my environment. I had presented 3 Vdisks to Red hat on the same & since it was not a production server, had them as Raid 0. We have total 91 disks on the SAN & one of them failed yesterday. My other Vdisks on the SAN which are presented to Windows with Raid 1 & 5 are allright but after replacing the faulty disk & rebuilding, Linux Vdisks entry on EVA has gone blank. On the Linux server, I can see them as mounted perhapes becuase I have not restarted this after disk crash & gives me I/O error when i try to read the files. I think if I restart the Linux, it might come with errors as it will not find all those mounts. Need some help on the steps to be taken to come back to original functioning state.

Regards,
Sandeep

[Eric van Dijken]

Unmount the partitions from the failed disk.
Replace the faulty EVA disk in the EVA.
Re-format the partitions on the failed vdisk.
Mount the partitions.
Restore your date from you back-up source (or reinstall)

[Mic V.]

Sandy, if your need is urgent, I suggest you may want to post to the Storage forum as well as here (Linux forum). Of course, calling HP is a good idea as well... :-)

Unfortunately, I have no EVA experience.

Mic

[Sandy_42]

Hi,

The partition is already unmounted on the Linux server & a new disk has been added. I did not understand "Re-format the partitions on the failed vdisk". After the disk was replaced & reconstructed, Linux partition was still in inoperative mode. Anyway, I make a Vdisk with Raid 5 & need to mount as separate mount all together. I am not sure how to make a LVM as that would have been better. Any steps for the same?

Regards,
Sandeep

[Eric van Dijken]

Sandy,

(Linux)
When your EVA Raid0 vdisk failed, you lost part (or greater part) of you disk partition information. From the linux point of view, your disk just became unreadable/corrupt, the replacement of the EVA disk, made the partition whole again, but the data on that part isn't what linux was/is expecting on it.

A re-format (mkfs -t ??? /dev/sd?) will fix this. After that you can remount the partitions and use them again.

(EVA)
Chances are very high that the EVA disabled/removed the Raid0 Vdisks as they used the "failed" disk. The EVA may have un-presented the vdisks that used the failed EVA disk. (On our site, we only use Vraid5 with Single Disk Fail protection) Try to present the linux vdisk again to the linux host, format and use it.

(LVM)
Not sure how LVM could have helped you with this, LVM is handy. But it will not handle a failed disk for you. Don't try to fix things on Linux, that the EVA can do for you.

[Sandy_42]

Hi,

I have made another Vdisk with Raid 5 of the same capacity for failed disk & presented to Linux host. Problem is as my Linux server has reboot & now it gets stuck at the volume name which got deleted because of disk crash. I deleted the entry from fstab for the mount point but it still looks for that partition entry. Not sure from where it gets loaded as I am not Linux expert. It was basically created as LVM of the SAN vdisk (2 vdisks of 100 + 50 GB, Raid 0).As it is not a production server, i can get away with that & than mount this new 150 GB Vraid 5 disk again if i somehhow know how to fix those mount errors at boot. Earlier lssd used to show me three entries as sda, sdb, sdc as I had 3 vdisks from SAN.

I need to know how to fix this boot mount error for the failed volume & than load the new vdisk for linux


Regards,
Sandy

[Eric van Dijken]

My best guess is that during system start-up, "vgscan" is run. This will find the LVM labels on the disks. And thus tries to re-enable the LVM volume. (But fails, and hangs)

I would unpresent the (old) raid0 vdisks, maybe even the (new) raid5 vdisks. And delete the vdisks completely on the EVA.

Create new raid5 vdisks and present them to you linux host. This will give you a fresh start.