I'm observing a bug in Client::setupClientLease() where the same LeasePinger instance gets scheduled twice resulting in failure.
Looking at the code
ret = invoker.invoke(new InvocationRequest(sessionId, subsystem, "$PING$", configuration, null, null));
if(ret instanceof InvocationResponse)
{
InvocationResponse resp = (InvocationResponse)ret;
Map respMap = resp.getPayload();
if(respMap != null)
{
Long leaseTimeoutValue = (Long)respMap.get("clientLeasePeriod");
long leaseTimeout = leaseTimeoutValue.longValue();
if(leaseTimeout > 0)
{
if(leasePinger == null)
{
leasePinger = new LeasePinger(this);
}
leasePinger.startPing(leaseTimeout);
}
}
}
The call to
leasePinger.startPing(leaseTimeout);
gets executed whether or not the leasePinger already exists.
We only want one pinger per client (I assume) so I have amended this in my local copy so startPing is only called in the case the leasePinger has just been created and not at any other time.
This seems to solve the problem.
