Solaris

Solaris 9

We had a problem server where only root was not mirrored (before my time). When I tried to mirror it, the live root slice bailed with errors at 97% so it couldn't be mirrored.

It's a matched pair of boxes (nfs1 and nfs2) and they are interchangeable with regards to the NFS functions. The slices are the same on all four disks (two mirrored disks in each system).

Since c1t0 on nfs2 was failing and we couldn't mirror it to c1t1, I broke the mirror on nfs1. We took c1t1 out of nfs2 and put it in c1t0's slot. We took c1t1 out of nfs1 and after changing the network information, /etc/system and /etc/vfstab, put it in c1t1 of nfs2.

The box booted fine and is accessible from the 'net. Because we wanted to save the data on c1t0, I modified vfstab to mount c1t1d0s0, c1t1d0s1, c1t0d0s3, c1t0d0s4, c1t0d0s5 and c1t0d0s6. The system rebooted again without a problem.

Problem:

Now I'm trying to mirror the system again. The problem is that I'm getting:

Code:

# metadb -f -a -c 2 c1t0d0s7 c1t1d0s7
metadb: nfs2: Bad address

I did some 'net surfing and didn't find anything that fit into my problem. Basically, there's a config file somewhere with the metainfo from nfs1 that needs to be cleared out.

I've looked at /kernel/drv/md.conf and even copied the file from c1t0. I checked /etc/lvm and removed the metastat info from md.cf since it was inaccurate.

I commented out the meta tools from /etc/inetd.conf and renamed the svm scripts so they wouldn't start.

On boot, I get a

Code:

metainit: nfs2: Bad address

so there's something somewhere that's still holding on to a nfs1 address.

Suggestions for other places to look?

Thanks.

Carl