mirroring vg00 on SAN (fc4700) taking FOREVER

[Joe Profaizer]

I have my boot disk on an fc4700 (fibre channel) path 1/10. I'm attempting to mirror the disk to another lun on path 0/4. The data is moving accross separate HBAs back to the LUN. My problem is that it is taking FOREVER versus internal SCSI drives.

Is there anyway to speed this process up? Let me know if you need anymore information. I'm on an N4000, using 1GB HBAs connectected via a brocade switch to the SAN.

Thanks,

...Joe

[Steven E. Protter]

This process should actually go pretty fast.

Are you using mirror/ux?

The next question is why? When connected to a large disk array, mirroring is usually best handled at the array.

You could be causing yourelf grief because the disk on the storage array is probably raid 5 nine bit parity and now you're tying to put mirror/ux raid 1 on top of that.

Might slow things up a bit. Worse than that, it might harm performance after the mirror is built.

My experience with mirror/ux is the mirrors build slowly but then are pretty snappy. Might want to do something else and just wait and see how the sytem performs later.

For boot disks, you might want to have your storage allocate a two whole disks for you to boot and mirror, without protection on the aray. If they are two different physical disks, mirror/ux will work on them, perhaps with less overhead.

I'd also check that the fiber cards are connected and functioning promerly. start with fcmsutil

SEP

[Joe Profaizer]

I mean it is crawling. It takes about 30 seconds to a minute for a stale PE to become current on a mirrored LV.

The fibre cards are operating fine as I used these paths for our oracle db (which is down now as I do the lvextend -m 1).

How harmful is it to abort the mirror disk process in the middle of execution? I'm not sure I can wait much longer.

[Joe Profaizer]

I think I may know what the problem is (other than I'm an idiot). Could I be mirroring to the same physical LUN. I'm basing this on that on path 0/4 the path is c1t0d0 and path 1/10 the disk is c5t0d0. I'm wondering if that's the same physical lun and is displayed by both HBAs as separate disks. For instance, when I create one LUN, both HBAs show a disk for the one LUN I created.

Could that be the problem?

[Jeff Schussele]

Hi Joe,

I don't think it will let you mirror to the same device via an alt path.
Open another session & do a vgdisplay -v on that VG & see if you actually see another device or just an alternate link.

Also definitely run
fcmsutil /dev/tdX stat -s
where X=td instance & look for error link losses & retransmits. Also should also verify your fibre. Could be a *really* crappy connection.

Rgds,
Jeff