Solaris Volume Manager naming schemes (Kilala.nl (Cailin Coilleach)) by Cailin Coilleach

When using SVM I've always been told to nicely match up meta device names with their corresponding slices on the hard disk. So a mirror called D0 would consist of meta devices d10 and d20 and would exists to mirror slice 0 of your main disk. Sounds logical right? Why would you needlessly confuse the matter?

Right now I'm working on a server whose boot disk should be rearranged due to space constraints. Looking at the SVM setup really makes me wonder what the person who built this box was drinking.

d0 -m d10 d20 1
d10 1 1 c1t0d0s0
d20 1 1 c1t1d0s0
d1 -m d11 d21 1
d11 1 1 c1t0d0s6
d21 1 1 c1t1d0s6
d2 -m d12 d22 1
d12 1 1 c1t0d0s1
d22 1 1 c1t1d0s1
d3 -m d13 d23 1
d13 1 1 c1t0d0s5
d23 1 1 c1t1d0s5
d4 -m d14 d24 1
d14 1 1 c1t0d0s7
d24 1 1 c1t1d0s7
d5 -m d15 d25 1
d15 1 1 c1t0d0s3
d25 1 1 c1t1d0s3

Ugh! *groan*

Posted in: meta devices , naming scheme , solaris , solaris volume manager

Why ZFS matters to the rest of us (Kilala.nl (Cailin Coilleach)) by Cailin Coilleach

Thanks to a link on the MacFreak fora I stumbled onto a great blog post explaining why ZFS is actually a big deal. The article approaches ZFS from the normal user's angle and actually did a good job explaining to me why I should care about ZFS.

Real nice stuff and I'm greatly looking forward to Mac OS X.5 which includes ZFS.

Posted in: mac os x , solaris , zfs

Crash course in new OSes (Kilala.nl (Cailin Coilleach)) by Cailin Coilleach

The past two weeks I've been learning new stuff at a very rapid pace, because my client uses only a few Solaris boxen and has no Linux whatsoever. So now I need to give myself a crash course in both AIX and Tru64 to do stuff that I used to do in a snap.

For example, there's adding a new SAN device to a box, so it can use it for a new file system. Luckily most of the steps that you need to take are the same on each platform. It's just that you need to use different commands and terms and that you can skip certain steps. The lists below show the instructions for creating a simple volume (no mirroring, striping, RAID tricks, whatever) on all three platforms.

Adding a new LUN to a Solaris box with SDS

1. Assign new LUN in the SAN fabric.
Not something I usually do.

2. Let the system search for new hardware.
devfsadm -C disks

3. Label the "disk".
format

4. Add the disk to the volume manager.
metainit -f $META 1 1 $DISK

5. Create a logical volume.
metainit $META -p $SOFTPART $SIZE

6. Create a filesystem
newfs /dev/md/$META

7. Test mount.
mount $MOUNT

8 Add to fstab.
vi /etc/vfstab


Adding a new LUN to an AIX box with LVM

1. Assign new LUN in the SAN fabric.
Not something I usually do.

2. Let the system search for new hardware.
cfgmgr

3. Label the "disk".
Not required on AIX.

4. Add the disk to a volume group.
mkvg -y $VOLGRP -s 64 -S $DISK

5. Create a logical volume.
mklv -y $VOLNAME -t jfs2 -c1 $VOLGRP $SIZE

6. Create a filesystem
crfs -v jfs2 -d '$VOLNAME' -m '$MOUNT' -p 'rw' -a agblksize='4096' -a logname='INLINE'

7. Test mount
mount $MOUNT

8 Add to fstab.
vi /etc/filesystems


Adding a new LUN to a Tru64 box running TruCluster

I'll edit this post to add these instructions tomorrow, or on Friday. I still need to try them out on a live box ;)


Anywho. It's all pretty damn interesting and it's a blast having to almost instantly know stuff that's completely new to me. An absolute challenge! It's also given me a bunch of eye openers!

For example I've always thought it natural that, in order to make a file system switch between nodes in your cluster, you'd have to jump through a bunch of hoops to make it happen. Well, not so with TruCluster! Here, you add the LUN, go through the hoops described above and that's it! The OS automagically takes care of the rest. That took my brain a few minutes to process ^_^

Posted in: aix , solaris , tru64 , volume management , work