Volume manager hosed?

[reinstalled]

It appears that all of my "playing around" with my mirrored OS
disks to get the right settings I have inadvertantly hosed volume mngr. I am trying to delete all mirroring that I had set up.
I know how to do this on a semi-working set but with this server when running any meta commands all I get back is:
/dev/md/admin: No such device or address

Is there a way to maualy delete all mirror info including the state replica's?

 

[bfitzmai]

You should have a soft link in /dev/md that looks something like /dev/md/admin -> /devices/pseudo/md@0:admin. This should have been set up once you rebooted the system after the metadevices were initialized.

 

[reinstalled]

Yep, that's exactly what it is. But it appers that it has lost any referrence to any part of my mirror set.
I get the No such device or address message using any meta command....metadetach, metaclear, metarepair, etc.....
I was thinking of reinstalling volume mngr over the top of the current installation but this isn't windows and I'm not sure that that will fix it. I'd rather know what caused the problem in the first place. My second server, which is a replica of the first, worked fine undoing everything and redoing it correctly.

 

[bfitzmai]

Look in the file /etc/system to see if you have meta entries for the md driver. Should look like: forceload misc:/md_mirror. If so, try a boot -r before you do any reinstallation.

 

[gkout]

Dont forget to change your /etc/vfstab file !!!

 

[reinstalled]

Well the forceload entries were not in /etc/system.
This is funny because when I had originaly set up mirroring those entries were ther. Also, when I went to look in /dev/md the md directory is gone. This is awfully strange to say the least. So, I think I need to re-install Volume Manager.
The only documentation I found that was somewhat specific says I need to install the SUNWCXall package cluster. Does this seem resonable?

 

[bfitzmai]

The forceload entries in /etc/system are created by command metaroot <metadevice>. This command will also modify the /etc/vfstab file so the root file system entry will use the metadevice instead of /dev/dsk. Did you run command metaroot /dev/dsk/c#t#d#s# when you were removing the mirrors a few days ago. It is my understanding that this will remove the forceload entries in /etc/system and change the root file system entry in /etc/vfstab.

 

[reinstalled]

OK that makes sense as to were the forceload entries went.
I used metaroot to specify /dev/dsk as the root device and must have removed thos entries. Still though, it leaves the question where did the md dir go?

 

[bfitzmai]

Did you run the metaclear command when you removed the mirrors? I have never had to remove mirrors once I have created them... So at this point, I am just guessing. I would think if you started your mirrors from scratch, creating the metadevices with metainit and running metaattach and metaroot, things should start working again.

 

[reinstalled]

No meta commands work. I beleive this is do to the md dir being missing.

 

[bfitzmai]

Yep reinstalled, I just checked some standalone Sunblades and all of them have the /dev/md directory. You may have to reinstall... From what I found, Solaris Volume Manager is part of the base system.

 

[reinstalled]

Well I guess I'll need to reinstall, I have never used the preserve existing data option during install. I ssh, ftp, and a couple other services set up in home as well as a bunch of scripts for our application. I assume that the preserve option will, in fact, preserve that data but will it also preserve the mount point and mount one it's own or do I need to re specify /home to mount manualy? Will I need to set up ftp again? and SSH?

 

[reinstalled]

Plus I have a Storedge array attached so I would need to copy out my sd.conf file as well I think.

All in all this will be a sketchy at best reinstall.

 

[bfitzmai]

FTP should be enabled by default... You will probably have to set up ssh... There is a ssh_config under /etc or /etc/ssh that you may want to perserve.