These steps are designed to restore one or more failed minions to an otherwise operational cluster. It is required that at least the master system is still functioning and that backup data for the failed minions is available. The servers that that restored minions will be installed on should be at the base OS level with no Cb Response software installed.
1. Stop cluster
2. Start postgres service
service cb-pgsql start
3. Determine node_id for missing minon(s) by querying postgres psql -p 5002 cb -c “select * from
4. Delete row from cluster_node_sensor_address table, where the nodeID is equal to the failed minion(s).
psql -p 5002 cb -c “delete from cluster_node_sensor_addresses where node_id = 2;”
*Where ‘2’ is the number from the missing minion
5. Stop postgres service
service cb-pgsql stop
6. Edit /etc/cb/cluster.conf file
7. In the [Cluster] section of the file, subtract N from NodeCount and NextSlaveAutoInc
lines, where N is the number of failed minions that are being restored
8. In the [Master] section of the file, add the the shard numbers from the ProcSolrShards line of the failed minions to the master’s ProcSolrShards in a comma separated list
9. Delete the [SlaveN ] section for each failed minion
10. Save the file
11. Add the minion(s) like you would as part of a normal install by running the following command and giving the appropriate IP or hostname and shard information
12. The master will install cb-enterprise on the minion system and allow for normal operation
13. Copy over backup data generated from the backup process and follow the restore procedures
14. Start cluster
These steps will be included in the next release of the 5.2 user guide.