Client pool doesn't recover after failure (client 1.3.11 / server 2.2.0)


#1

When I fail a node, the clients continue happily until I add the node back and rebalance.
About halfway through the rebalance, some nodes will disconnect and never reconnect.

The clients are established in a static method on application start. Any insight into what’s going on when they fail to recover, and what to do about it?


#2

@jmcgill_public -

Can you post client logs?


#3

Okay to attach here? Or I can dropbox?


#4

You can create an NCBC and attach them here: https://issues.couchbase.com/browse/NCBC

Have you considered updating to the v2.1.0 of the SDK? I suggest you do.

-Jeff


#5

We have two problems upgrading to 2.1.x, first is that we can’t bring the project into .NET 4.5 (It’s a very big set of projects, and couchbase caching is a tiny piece of it), and second is that the 2.1.x API seems wildly different, starting with the lack of a CouchbaseClient object.

I’m working on a test case that will generate logs to show what’s happening and I’ll put it in the jira (NCBC).


#6

This is true, 2.1.0 requires .NET 4.5 or greater and the API has changed; there is a ClusterHelper object that can help with the integration, however, it’s in the main Couchbase namespace.


#7

The level of effort behind migrating a very large project to .NET 4.5 is prohibitive. The API changes would not be too hard, but it’s surprisingly difficult to migrate the project.