As we known, when a node down, the cluster need somes time to auto-failover it.
during this period, some request will be serviced, but some requests may be failed.
my question is how to handle those failed request to make it be serviced?
There is a replica read command that you can use, but it is not yet implemented in all of our SDK's. Which SDK are you using?
Thanks for your reply. my SDK is Java couchbase-client-1.0.2.
But i think i didn't describe the situation clearly. Following are deitals:
Our env has CouchbaseNodeA, CouchbaseNodeB, CouchbaseNodeC as a cluster,
this cluster has been configure as auto-failover with 30 seconds.
our SDK is java client 1.0.2; using apache common pool to create a couchbase connection pool.
firstly, we do some operation (save/get data from couchbase) , it all works well;
then , we shutdown a cluster node,
as we known, the cluster need some time to auto failover the node,
during the time, some operation on the cluster will be failed, but some ops can be serviced.
after auto-failover, everything works.
so my question is that Do we have any way to make every operations successful
during the time cluster auto failover the down node?
No. If a node is down you will not be able to get any data from that node. One thing you could do though is first do a get and if it fails do a replica read to get the data from a replica node. You are however unable to create or update data on a replica node. I don't think replica read support will be in the next release of the java sdk, but it will definitely be in the following one.
© 2013 COUCHBASE All rights reserved.