Velocity has been renamed to AppFabric Caching Services. I've taken the suggested updates from this stackoverflow article: http://stackoverflow.com/questions/3233792/appfabric-could-not-contact-the-cache-service and improved them.
lock/unlock are fixed.
exceptions are still swallowed but are now logged
item will now remain locked after a Put() operation if it was locked already - see http://social.msdn.microsoft.com/Forums/en/velocity/thread/2b6df3a7-0bf8-4386-ba2d-884fb1a29e0e for explanation
if cache cluster is restarted, we will receive an error "ErrorCode<ERRCA0005>:SubStatus<ES0001>:Region referred to does not exist. Use CreateRegion API to fix the error". Open to options on how to resolve that.
Please note: as part of applying this patch, the following files may be removed from lib\net\3.5:
Attaching new version of patch. This version wraps the DataCache operations so that if the cache cluster is restarted, it will trap the RegionDoesNotExist error and recreate the region.
I appreciate there is cache provider for AppFavric Cache, but I think that the usage of locks dictionary in VelocityClient is problematic due to following reasons:
The client class is not thread safe (according to ICache comments should be) as distinct Add, Remove, Count may be performed concurrenctly. This could lead to race condions and unexpected dictionary exceptions.
(ii) If we are in cluster, each node would have its own locks dictionary, which could again lead to potential unlock by put.