I got the same result
I Don't understand this error!
Boot from CD into single-user.
# mount /dev/dsk/c0t0d0s0 /a
# cd /a
# ls -l
Does the /devices directory exist? If not, create it
# mkdir devices
Try boot with reconfigure again.
Perhaps post the output of the 'ls -l' so we can see if it looks right.
1 Like
# ls -l
not 'ls -i'
sorry about that
But it shows that devices is not present.
Create it with
mkdir devices
and then reboot with
reboot -- -r
and hope the -r (reconfigure) will create them all.
3 Likes
Agreed. Let's give that a go.
It wasn't there.
I created devices and made sure it was there.
Init 0
Boot disk -r
And the result is below
It hangs in this state
Well you can do the same and create /proc
However, for whatever reason it seems that some files in /etc are also missing, but try creating /proc and rebooting (with -r).
Whilst you are creating /proc, go into /etc and see whether the files it's complaining are missing are, indeed, not there.
1 Like
I didn't get any error with CPIO in this last attempt.
ok, I will create it
Perhaps take a look at the good machine and see if those files exist over there. I think that they should do.
Could also list the cpio archive to see what you have. Could redirect output to a file and then grep (but we might need to explain that to you?)
1 Like
I think the find options were wrong in find ... | cpio ...
The -fstype ufs has the mount points excluded because they belong to different fs types: procfs, mntfs, tmpfs, devfs, ....
Perhaps a better command would have been
find . -type d ! -fstype ufs -prune -o -print | cpio -ov -O /mnt/CPIO/backup.cpio
Now you must create the missing mount points:
cd /a
mkdir devices proc system/object etc/svc/volatile
touch etc/mnttab
2 Likes
Yes, but /root, /var and /usr are certainly ufs.
I thought of !-fstype proc plus !-fstype nfs but in the end didn't recommend that.
Anyway, surely, -fstype ufs shouldn't exclude anything in /etc, should it?
@MadeInGermany please feel free to specify a cpio archive command line.
There ain't so much data so it won't take long to run.
We could then stream that in over the top of what we've already got.
Alternatively, we could just cpio archive the /etc directory from the good machine and restore that over the top (to insert the missing files).
To me, this is getting pretty close.
Question is: Why are some /etc files missing?????
Of course, another way, would be to fssnap/ufsdump the 3 filesystems from the good machine and ufsrestore them onto the clone box. We ain't finished yet!
1 Like
I have edited my last post and put a find command. Hope it's correct now.
Yes, ufsdump/ufsrestore is perfect, because ufsdump runs on the raw device and does include the "overmounted" UFS mount points. But each file system needs a separate ufsdump/ufsrestore.
How about that now?
I have created /system/contract as well, since it was missing.
Anyway, what is the status now?
I'm currently in single user mode.
Can I launch the CDE?
Shall I go back and redo the find options?
Waiting for your feedback
Edit: I tried to reboot again...so, I'm getting a prompt " The /tmp file system (/dev/rdsk/c0t0d0s4) is being checked."
Warning-unable to repair the /tmp filesystem. run fsk mannually (fsk -F ufs /dev/rdsk/c0t0d0s4).
Does this have to do anything with the fact that we allocated zero cylinders for Slice 4?
1 Like
That tells me that slice 4 should be a filesystem /tmp which wasn't named in the VTOC as such. /tmp is not normally a separate filesystem so perhaps we can edit it out of /etc/vfstab and let it be integrated with root. However, that wouldn't be an exact clone of the system.
So option is to do it all again and give slices 4 & 5 some space allocation and actually name slice 4 'tmp', and create a filesystem there too 'newfs' it.
However, I'd still like to know why there are files missing from /etc. when we archived the whole system.
1 Like
A graphical desktop like CDE cannot be run in single-user mode.
You are currently in an emergency single-user mode that is slightly different from a wanted single-user mode.
Normally /tmp is tmpfs (a in-memory filesystem). But can be UFS as well.
Please show
cat /etc/vfstab
and
prtvtoc
1 Like
I'm aware of that. I meant can I boot in multi-user mode and launch the desktop.
It was forced to go into emergency single-user mode because of the above warning.
No, you must fix the emergency problem first. Then it will boot to multi-user.
1 Like
You could make sure that /tmp exists on the root filesystem (exactly what you did with /devices and /proc, and then comment out the entry in /etc/vfstab for /tmp.
Then the system will think that tmp files are stored in the /tmp and not try to fsck a filesystem at boot time.
1 Like