Uncaught exception in enqueueOperation

[jrray]

The calls to mconn.enqueueOperation in SpyMemcachedIntegration can throw IllegalStateException when the queue is full:

java.lang.IllegalStateException: Timed out waiting to add Cmd: 1 Opaque: 229348 Key: XXX Cas: 0 Exp: 1556244278 Flags: 0 Data Length: 0(max wait=10000ms)
        at net.spy.memcached.protocol.TCPMemcachedNodeImpl.addOp(TCPMemcachedNodeImpl.java:362)
        at net.spy.memcached.MemcachedConnection.addOperation(MemcachedConnection.java:1278)
        at net.spy.memcached.MemcachedConnection.addOperation(MemcachedConnection.java:1240)
        at net.spy.memcached.MemcachedConnection.enqueueOperation(MemcachedConnection.java:1196)
        at shade.memcached.internals.SpyMemcachedIntegration.realAsyncSet(SpyMemcachedIntegration.scala:235)
        at shade.memcached.MemcachedImpl.set(MemcachedImpl.scala:60)

This ends up breaking the implied contract of always returning a Future when calling set, etc.

[lloydmeta]

Oh nice find. Maybe we can wrap those queue inserts in a try {} catch {}, maybe abstracted in another method, and on failure, fulfil the Promise with the exception.

[jrray]

My "random guy on the internet" two cents would be push the call to mconn.enqueueOperation down into prepareFuture where this error case can be handled in a central place. It looks like every call to prepareFuture(key, op, ...) is preceded by a call to mcomm.enqueueOperation(key, op).