The 1.1.7 release is the seventh bugfix release for the 1.1 series. This release mostly brings better stability during rebalance and failover phases. Also, fail-fast exceptions have been implemented to disallow certain operations on memcache buckets.
For lots of HTTP responses (Views, DesignDocuments) we now retry until timeout when its not a 200 success. This especially helps for transient errors and server redirects. This should bring better stability for http requests during the rebalance process.
The SDK now "fails fast" when one tries to use persistence options (ReplicateTo, PersistTo) with memcached-type buckets. This is not supported and now immediately throws a exception instead of failing further down the stack.
Spymemcached has been upgraded to 2.9.0. One of the major enhancements is support for SLF4J, which is now also available for the Java SDK.
In cases where there is no master partition available for the key (so in the internal vbucket map its ID is -1), now we correctly cancel the operation and also trigger reconfiguration (to make sure the system eventually recovers).
the IO thread are caught and we are making sure that those type
of exceptions dont terminate it. There are rare occurences where
they can occur internally, but they are temporary. Note that
they still get logged properly.
It can be possible that View operations don't recover properly through rebalance operations when the node on which the client is currently receiving configuration updates from is removed. This will be addressed in one of the next bugfix releases.