Galera

Restore MYSQL cluster

1. Kill all mysql process on whole cluster (check via ps aux| grep mysql)
2. Remove grastate.data and ib_logfiles
   • rm /var/lib/mysql/grastate.data
   • rm /var/lib/mysql/ib_logfile*
3. Edit this row in /etc/mysql/my.cnf on `"wsrep_cluster_address="gcomm://""``
4. Start MySql on this node
   • validate status of MySql cluster salt-call mysql.status | grep -A1 wsrep_cluster_size
5. Start MySql on another one node
6. Wait when second rejoin cluster, when is cluster = 2 start MySql process on last node
7. salt-call mysql.status | grep -A1 wsrep_cluster_size cluster size should be equal to 3
   • re-run galera state on first

Offline migration between two compute nodes

1. Stop the instance
2. Rsync qcow image to right destination rsync -Pa /var/lib/nova/instances/instance-000004f2 destination_hostname:/var/lib/nova/instances/
3. Update record in MySql database UPDATE instances SET node='os-kvm-prod-dmz-self014.mgm.avg.com',host='os-kvm-prod-dmz-self014' WHERE uuid ='9f2d63b5-122c-47b1-812a-1bec3d40fbb9';

More Details

Detach Volume

1. Log in to database
2. use cinder table
3. Update correct record
   • UPDATEcinder.volumesSETstatus='available',attach_status='detached' WHEREid`='2d66d83b-a644-480a-a457-56075838612b';``

cinder service-disable cpt06@EasyTier cinder-volume a delete from db: mysql -uroot -pcloudlab -e "update services set deleted = 1 where host like 'cpt06@EasyTier' and disabled = 1 " cinder

MySql Commands

show processlist - It shows on witch node is active connections show status - In this table is lot of interesting informations.