V210 Disk Mirroring

We have a new V210 with 2x1.3GHz CPU, 2GB RAM and 2x73GB HDD.

It came from Sun's rental company with Solaris 9 installed, as requested, and the following partitions on disk 0:

0 - root 4GB

1 - swap 4GB

3 - /var 4GB

4 - unused

5 - unused

6 - unused

7 - /export/home 56GB

Our aim is to mirror root, swap, /var and /export/home between disk 0 and disk 1 using SVM.

In order to setup the state database replicas, 500MB space was allocated to partition 6 by using disk/format/partition to reduce the size of partition 7. Once complete disk 1 was setup the same as disk 0.

The meta devices for root (d0), swap (d1) and /var (d3) were then created and are all working fine and dandy.

However, we can't mount metadevice d7 on /export/home, even though the metadevices d7, d17 and d27 are reporting as 'okay' using metastat. We get an 'IO error' message when a mount is attempted.

My guess is that re-sizing partition c0t0d0s7 has caused problems.

Can anyone help?

[1063 byte] By [moyes_boy] at [2007-11-25 22:46:40]
# 1
If you did not newfs /dev/rdsk/c0t0d0s7 and then restore after "borrowing " 500MB and labeling then that is where you I/O error is coming from.
jds2n at 2007-7-5 17:02:01 > top of Java-index,Sun Hardware,Servers - General Discussion...
# 2

Additionally to Johns comments, the normal slice to take space from for metadb slices is swap. Solaris can use 0.5 to 1.5 times the systems RAM for swap space. Typically the disk layout for a system that is to have a root mirror is:

0 /

1 swap

3 /var ( + 2GB )

4 dedicated for crash dumps ( + 2GB )

5 whatever ( /export/home )

6 whatever ( /export/home ) or metadb slice 30MB

7 metadb slice 30MB

mlennon at 2007-7-5 17:02:01 > top of Java-index,Sun Hardware,Servers - General Discussion...
# 3

Thanks for the pointers.

We've done a newfs on c0t0d0s7 and c0t1d0s7, run fsck and re-built the mirrors. All seems well so far.

The main lesson we've learnt here is to setup the mirrors before we start installing software and dropping files into /export/home. Doh, etc!

Many thanks :-)

moyes_boy at 2007-7-5 17:02:01 > top of Java-index,Sun Hardware,Servers - General Discussion...
# 4

Looks like I spoke too soon :-(

We can run fsck OK against all of the filesystems, and we can see the filesystem mounted in the external T3 array, but the server keeps dropping back to the ok prompt with the following message:

"SC Alert: SC Request to send Break to host."

Are there any tests that we can run before we call the support line?

Thanks,

Paul.

moyes_boy at 2007-7-5 17:02:01 > top of Java-index,Sun Hardware,Servers - General Discussion...
# 5
Did you perform a metaroot command?I don't think that is the problem though. Let me know what you find out from Sun.
dimndskier at 2007-7-5 17:02:01 > top of Java-index,Sun Hardware,Servers - General Discussion...
# 6
The attached Windows PC running hyper terminal as a console to the server was the problem. As soon as we disconnected the keyboard the problem went away.Paul.
moyes_boy at 2007-7-5 17:02:01 > top of Java-index,Sun Hardware,Servers - General Discussion...