[Systems] Freedom disk/partition layout

[Bernie Innocenti]

On 01/08/16 01:07, Samuel Cantero wrote:
> Hey guys,
> 
> I was checking freedom disk/partition layout. I've found:
> 
> 1) We have 4 disks. 2 disks are being used by Sugar and 2 by OLE.
> 2) The disks sdc and sdd used by Sugar are partitioned in the following way:
>    a) three 10 GB partitions (partitions 1-3).
>    b) one 1970 GB partition (partition 4).
> 3) We have five RAID1 arrays:
>    a) /dev/md0: sdc1 + sdd1. Currently used by /.
>    b) /dev/md1: sdc2 + sdd2. Not used.
>    c) /dev/md2: sdc3 + sdd3. Not used.
>    d) /dev/md3: sdc4 + sdd4. Physical Volume for the VG freedom-lvm.
>    e) /dev/md4: sda1 + sdb1. Physical Volume for the VG ole.
> 
> 4) In freedom-lvm where we have the following Logical Volumes (LVs):
>    a) backup => 1 TB.
>    b) docker_extra_storage => 15 GB. Not enough. It is using btrfs and
> the amount of space used differs from df -h result. We should use the
> Docker device mapper instead.
>    c) freedom-virtual => 500 GB. qcow files for a bunch of VMs. Not sure
> which VMs are being used this space. The VMs currently running are:
> ole, hanginggarden, chat, munin, kuckuck, hammock, beacon, owncloud,
> pirate. I guess all of these VMs belong to Dogi. There are also qcow
> images for apparently decommissioned VMS.
>    d) hammock-data => 150 GB. 
>    e) hanginggarden-data => 50 GB.
>    f) ole-data => 100 GB.
>    g) socialhelp => 20 GB. This wrong named LV is used to host some SL
> stuff.
>  
> 5) In ole there is only one LV named hammock with 2.73 TB.
> 
> Bernie bought two 2 TB (~1.8TB) new disks for SL the other day. We are
> planning to create the partitions for the new disks and rsync the
> current content.
> 
> I suggest the following layout:
> 
> (a) md0: sdx1 [10GB] + sdy1 [10GB] => /

10GB has become a bit tight for the root, let's make it 20GB.

> (b) md1: sdx2 [500GB] + sdy2 [500GB] => PV for Docker.
> (c) md2: sdx3 [rest] + sdy3 [rest] => PV for the backup (1 TB) and srv
> partiton (replace socialhelp).


Is it not possible to use one big volume group for both docker and the
other partitions?

Once we've split the disk into separate raid arrays, reshaping it will
require extensive downtime.


> The idea is to use the the docker DM [1] instead of btrfs. We should
> also deploy kubernetes [2] or something similar in order to improve our
> container infra.
> 
> Any suggestions?
>
> Best regards,
> 
> Samuel C.
> 
> [1]
> https://docs.docker.com/engine/userguide/storagedriver/device-mapper-driver/ 
> [2] http://kubernetes.io/

-- 
 _ // Bernie Innocenti
 \X/  http://codewiz.org