Just attempted an upgrade on a 3-node cluster. Community edition, centos 6.6.
The plan was to do an online rolling swap. Remove a node, rebalance, shut down the node’s services, remove the rpm, install new rpm, re-add node to cluster. Repeat.
Some questions and notes on the upgrade.
-
Installing the new rpm produces the error/warning:
1:couchbase-server-commun########################################### [100%] tar: /opt/apps/couchbase/lib/python/pysqlite2.tar: Cannot open: No such file or directory tar: Error is not recoverable: exiting now Starting couchbase-server-community I checked and no such tar file exists in that directory nor in the rpm itself My rpm command is: rpm -ivh --relocate /opt/couchbase=/opt/apps/couchbase couchbase-server-community-4.0.0-centos6.x86_64.rpm -
Re-adding a node (new version) to the cluster (old version) produces the error:
Attention - Node attempting to join an older cluster. Some of the selected services are not available.
I assume this is simply because the services (data/index/query) are separated out now. Choosing just the “data” service allowed me to continue. This meant after getting all nodes on the new version, they all had just the “data” service.
To get them to have all services, do i have to go back to each node and do a remove/rebalance/re-add and then i’ll be able to pick all the services on setup?
- to use N1QL, should i have the Query service running on ALL of my nodes? And is there no way of using the Query service without running the Index service? In the setup it auto-checks the Index service if I check the Query service.
Thanks,
-Tony