Lost a lot of data where a node crash

And the client sdk still send the data to the crash nodes, It can not select another usable node to send.

my Test as follows:
I have start 4 nodes for the coucbase cluster.
1.start a client (1.4.0) to set 100 0000 items to the couchbase!
2.before the client add completed I kill one of the node.
3.and the client logs as "Discarding partially completed op: Cmd: 2 Opaque: 72280 Key: key_72279 Cas: 0 Exp: 0 Flags: 0 Data Length:"
4.then when the client complete for send the 100 0000 items, the couchbase will lost 25 0000 items just
the 1/4 of all the data.

how to solve this problem anyone have any suggestion would be help.



I have no idea about your code, but you need to make sure to check on the return values, either by blocking or through a listener. If the op fails, this will be noted in the operation status so you can retry from your application level or fail appropriately.

Can you show some code?