Upgrading oyster.
This is the plan for upgrading the OS and the memory and the disk partitions on oyster simultaneously.
-
Pacify nagios for oyster.
-
Pacify nagios for its MPU guests too.
-
Migrate essential VMs elsewhere.
- Stop oyster's VMs from suspending and resuming at every reboot:
!boot.services mREMOVE(rc_libvirtguests)
- Shut down every VM in
rvirsh oyster list
:
kvmtool shutdown --name rembrandt
kvmtool shutdown --name nimrodkvm
kvmtool shutdown --name tetrakvm
kvmtool shutdown --name praecox
kvmtool shutdown --name knox
kvmtool shutdown --name sanremo
kvmtool shutdown --name tench
kvmtool shutdown --name nescxldb
kvmtool shutdown --name barb
kvmtool shutdown --name tarnkvm
kvmtool shutdown --name natter
kvmtool shutdown --name charlar
kvmtool shutdown --name canary
kvmtool shutdown --name redmires
kvmtool shutdown --name yubtest
kvmtool shutdown --name dstwo
kvmtool shutdown --name dsone
kvmtool shutdown --name guppy
kvmtool shutdown --name bakerstreet
kvmtool shutdown --name huxley
kvmtool shutdown --name dunkvm
kvmtool shutdown --name banter
kvmtool shutdown --name craic
kvmtool shutdown --name lardup
kvmtool shutdown --name gabble
kvmtool shutdown --name heda
kvmtool shutdown --name rhacotis
kvmtool shutdown --name arachne
kvmtool shutdown --name storck
- Check
rvirsh oyster list
to make sure they're all down.
- Check that
/var/lib/libvirt/qemu
is empty (apart from 3 empty directories).
-
rfe lcfg/oyster
and remove DICE_STICK_WITH_SL65
- Wait for the new profile to hit oyster and check
/var/lcfg/log/client
etc.
- Install any new packages before rebooting, to minimise later reboots:
om updaterpms run -- -f
- Shut oyster down.
- Install the new memory.
- Switch oyster on.
- Ensure that the new memory is recognised and passed by the POST.
- If possible boot to single user.
- Get the new partition's UUID:
/sbin/blkid /dev/sda4 -o value -s UUID
- Edit sda4 and its UUID into
/var/lcfg/conf/fstab/fstab.sda
-
om fstab configure
or edit sda4 and its UUID into /etc/fstab
-
mount -a
to check the new partition.
- Go to multi-user or reboot.
- Start the VMs:
/etc/init.d/libvirt-guests start
- Or if that doesn't work when VMs aren't running at all, start them up individually:
kvmtool start --name rembrandt
kvmtool start --name nimrodkvm
kvmtool start --name tetrakvm
kvmtool start --name praecox
kvmtool start --name knox
kvmtool start --name sanremo
kvmtool start --name tench
kvmtool start --name nescxldb
kvmtool start --name barb
kvmtool start --name tarnkvm
kvmtool start --name natter
kvmtool start --name charlar
kvmtool start --name canary
kvmtool start --name redmires
kvmtool start --name yubtest
kvmtool start --name dstwo
kvmtool start --name dsone
kvmtool start --name guppy
kvmtool start --name bakerstreet
kvmtool start --name huxley
kvmtool start --name dunkvm
kvmtool start --name banter
kvmtool start --name craic
kvmtool start --name lardup
kvmtool start --name gabble
kvmtool start --name heda
kvmtool start --name rhacotis
kvmtool start --name arachne
kvmtool start --name storck
- Remove the oyster details from https://computing.help.inf.ed.ac.uk/scheduled-downtime
- Migrate the essential VMs back from hammersmith and jubilee.
- Remove the temporary storage and symlinks from hammersmith and jubilee.
Topic revision: r1 - 31 Aug 2015 - 14:18:31 -
ChrisCooke