Solaris

I have a disk in maintanance status . I executed metareplace -e partition device . It resynced upto 5 % and then stopped . In the /var/adm/messages it showed the write error .

My question is if I analyse and repair the disk in format command , will my write error go . I dont have 0 hard and soft error in isostat -En . Is there any possibility of data loss .

maybe you can further investigate by :
#metadb -i

The disk checks via format should not cause any data loss, so long as you stay clear from any destructive (write) tests. Check the disk for any bad blocks and see if you can sync the mirrors if no disk issues are reported. If you still cannot re-sync, I'd bite the bullet and replace the disk - better to be safe than sorry, especially if it's in a production environment.

Replace the drive now; test/format later.

Not critical but prevents more spurious errors: If you want to do any destructive tests, metadetach the failing device from your master metadevice first.

Otherwise, to paraphrase what 'seg' says: Your disk is buggered, replace it now. Even if you get it going now, it's going to fail again really soon.

edit: Oh and I wouldn't worry about data loss at the moment, if you have something to resync against, then that disk will still have a good copy of your data.

Sounds like the disk is hosed, though it may be worthwhile to do an analyze refresh on the disk, its non destructive and checks reads and writes each block to/from the disk.. It will at least show up if there are issues before you replace the thing ...

If you have a mirror with good data you wont lose any data ....

It happened with me before. Try to metadetach, metaclear, then metainit again, re-create it and it might work

However SUN says whenever the status is "need maintenance" then replace the disk. Better safe than sorry