SCI-CD

h1. Adding extra volume group

If your server has two types of disks, for example 2xSATA + 8xSAS, then you need to initialize the second VG and (if needed) to make it as a main for creating of virtual machimes:

This operations should be done on both nodes:

Partition disks:

<pre>

fdisk /dev/sdc

</pre>

make a primary partition on whole disk

If you use Software RAID, then copy this prtitioning to all the disks (be careful, don't mix up the disks):

<pre>

for i in d e f g h i j; do sfdisk -d /dev/sdc|sfdisk /dev/sd$i; done

</pre>

Initialize md, vg (skip mdadm command if you have hardware RAID)

<pre>

mdadm --create -n8 -l10 /dev/md3 /dev/sdc1 /dev/sdd1 /dev/sde1 /dev/sdf1 /dev/sdg1 /dev/sdh1 /dev/sdi1 /dev/sdj1

pvcreate /dev/md3

vgcreate sas /dev/md3

</pre>

And, finally, tune the cluster defaults to use the new vg (issue this one time on the master node):

<pre>

gnt-cluster modify --vg-name=sas --disk-parameters=drbd:metavg=sas

</pre>

By default all disks have set the deadline scheduler.

If you use the hardware RAID or separate external storage, connected via FC/Infiniband/iSCSI, then, POSSIBLE, will be better to use noop scheduler for the underlying physical disks.

Nevertheless, the deadline scheduler will be sufficient for all disk types.

The article about comparing the three variants of offset in raid10 (SCI use the default one, NEAR):

http://www.ilsistemista.net/index.php/linux-a-unix/35-linux-software-raid-10-layouts-performance-near-far-and-offset-benchmark-analysis.html?start=1