Looks like this was fixed in 3.5 as a result of ZOOKEEPER-1505. Perhaps we
can grab something from that (larger) patch to fix it. I agree that it
looks like a bug.
On Fri, Jul 26, 2013 at 12:59 AM, Ted Dunning <[EMAIL PROTECTED]> wrote:
> That does look wrong at first glance. Certainly the hang is a bad thing,
> no matter the cause.
> Switching hash map types is low impact, but do you have any way to
> replicate the problem so that we can tell if it really fixes things?
> On Thu, Jul 25, 2013 at 9:52 PM, Dave Latham <[EMAIL PROTECTED]> wrote:
> > We bumped into a what looks like a pretty serious but simple concurrency
> > bug in ZOOKEEPER-1731. It looks like the fix may be as simple as using a
> > ConcurrentHashMap rather than a HashMap. Is anyone able to take a quick
> > look?
> > Thanks,
> > Dave