vm using z2-4 as primary :
- aviosys.binbang.vm.gnt
- gnt2.vm.gnt
- harvest.vm.gnt
- mediagateusa.binbang.vm.gnt
- munin.fsffrance.vm.gnt
- neufbox.binbang.vm.gnt
- openvpn.fsffrance.vm.gnt
- pokme.vm.gnt
No additional ip (87.98.156.150 is not used anymore). Binbang vm can be down, gnt2.vm.gnt too (test vm).
- munin.fsffrance.vm.gnt : use proxy on z2-2
- openvpn.fsffrance.vm.gnt : not used atm
- harvest.vm.gnt : use proxy on z2-2
- pokme.vm.gnt : not used
vm using z2-4 as secondary :
| Instance | Primary node |
| booken.binbang.vm.gnt | z2-1.host.gnt |
| dmail2.dachary.vm.gnt | z2-5.host.gnt |
| hanabi.mekensleep.vm.gnt | z2-6.host.gnt |
| mvixusa.binbang.vm.gnt | z2-3.host.gnt |
| neufbox-fc.binbang.vm.gnt | z2-5.host.gnt |
| ns1.fsffrance.vm.gnt | z2-5.host.gnt |
| proxy.mekensleep.vm.gnt | z2-6.host.gnt |
| savannah.vm.gnt | z2-7.host.gnt |
| silva.vm.gnt | z2-5.host.gnt |
| speedtouch-716g.binbang.vm.gnt | z2-2.host.gnt |
| trac.binbang.vm.gnt | z2-3.host.gnt |
| trac.fsffrance.vm.gnt | z2-1.host.gnt |
| vpn-dns.vm.gnt | z2-5.host.gnt |
| wetball.mekensleep.vm.gnt | z2-6.host.gnt |
| yocto.vm.gnt | z2-5.host.gnt |
vms using z2-[123] as primary should also use one of them as secondary.
Migration
0) backup z2-4 on rosiers
Active vms
1) shutdown binbang vms on z2-4
2) failover all vms to their secondary node. Because they use z2-2 proxy, there is no ip / config to move. Routes will be updated by ospf.
3) check they are still available after failover
Migrate secondaries
0) down all binbang vm except binbang.vm.gnt and trac.binbang.vm.gnt if not done
1) Relocate all secondaries :
| VM | new secondary |
| aviosys.binbang.vm.gnt | z2-1.host.gnt |
| gnt2.vm.gnt | z2-5.host.gnt |
| harvest.vm.gnt | z2-2.host.gnt |
| mediagateusa.binbang.vm.gnt | z2-6.host.gnt |
| munin.fsffrance.vm.gnt | z2-6.host.gnt |
| neufbox.binbang.vm.gnt | z2-3.host.gnt |
| openvpn.fsffrance.vm.gnt | z2-6.host.gnt |
| pokme.vm.gnt | z2-6.host.gnt |
| booken.binbang.vm.gnt | z2-3.host.gnt |
| dmail2.dachary.vm.gnt | z2-6.host.gnt |
| hanabi.mekensleep.vm.gnt | z2-5.host.gnt |
| mvixusa.binbang.vm.gnt | z2-1.host.gnt |
| neufbox-fc.binbang.vm.gnt | z2-6.host.gnt |
| ns1.fsffrance.vm.gnt | z2-6.host.gnt |
| proxy.mekensleep.vm.gnt | z2-5.host.gnt |
| savannah.vm.gnt | z2-6.host.gnt |
| silva.vm.gnt | z2-7.host.gnt |
| speedtouch-716g.binbang.vm.gnt | z2-3.host.gnt |
| trac.binbang.vm.gnt | z2-1.host.gnt |
| trac.fsffrance.vm.gnt | z2-2.host.gnt |
| vpn-dns.vm.gnt | z2-6.host.gnt |
| wetball.mekensleep.vm.gnt | z2-5.host.gnt |
| yocto.vm.gnt | z2-6.host.gnt |
using :
gnt-instance replace-disk --submit -n z2-X.host.gnt vm_name
the --submit flag allows to queue all requests.
2) wait for all syncs to be finished, z2-4 should then be empty (chec with gnt-node info z2-4.host.gnt)
3) remove the node from the cluster :
gnt-cluster remove z2-4.host.gnt
====
0) backup done on z2-7, lv z2-4-2009-11
1)
z2-2:~# gnt-instance failover openvpn.fsffrance.vm.gnt z2-2:~# gnt-instance failover munin.fsffrance.vm.gnt z2-2:~# gnt-instance failover harvest.vm.gnt z2-2:~# gnt-instance failover pokme.vm.gnt
migration
After z2-2 crash + ganeti bug when tried a gnt-cluster repair-disk-size, (see http://code.google.com/p/ganeti/issues/detail?id=80 ), all the vm are back to normal and can be migrated one after the other.
