The Enyim client is the only cllent hitting Membase, so somehow moxy is
being used Perry, possibly because we are using a memcached type client ?.
Something went wrong when we switched the Enyim client from membase to
memcached type. The bucket is memcached type as well. I logged over a
million errors in 2 hours, and these moxi process failures started happening
at the same time server side.
If you follow my trail on this group you can see we reverted the client to
type 1 because of unexplained 401 errors coming back from the pool config
url when Enyim tries to reconfigure the socket pool. This seems due to bad
basic auth header being sent by the client, even though we never configured
any credentials to send. That along with socket timeouts (despite 20 second
timeout limits) and other stuff we just never could explain. I've worked
hard to make this fault tolerant, adding the throttling policy client side
to not shut down the pool when occasional errors happen. So far, it's been a
nightmare to be honest. So many issues showed up only under load, which is
very hard to test reliably and prepare for. We've had a lot of application
errors and fallout from this.
Just trying to give you some feedback you can use, and help others trying to
get this going. I don't have time to mess with it anymore. Thanks for all