Hi,
seeing this here as well. Basically at some point during a repair within
the cluster (incremental, parallel, partitioner range, only one node at
a time) some node (sometimes the repairing node, sometimes another)
starts piling up READs, i.e. pending reads go through the roof. Despite
only one
Hi,
we just had such an incident again, and this time we were able to grab a
jstack-threaddump.
One thread from this stackdump actually shows a ReadVerbHandler calling the
RateLimiter. From my understanding this should not be:
Thread 30346: (state = BLOCKED)
- java.lang.Thread.sleep(long) @bci=
Hi,
I just found another one. Its basically the same, but I'll post it anyway:
Thread 84311: (state = BLOCKED)
- java.lang.Thread.sleep(long) @bci=0 (Compiled frame; information may be
imprecise)
- java.lang.Thread.sleep(long, int) @bci=57, line=340 (Compiled frame)
- java.util.concurrent.Time
Yep, definitely a bug. Introduced by CASSANDRA-9240 (me; mea culpa).
I've filed a JIRA for you: CASSANDRA-11301
On 3 March 2016 at 14:10, horschi wrote:
> Hi,
>
> I just found another one. Its basically the same, but I'll post it anyway:
>
> Thread 84311: (state = BLOCKED)
> - java.lang.Thread
Thanks. That's good to know.
On Thu, Mar 3, 2016 at 8:48 PM, Benedict Elliott Smith
wrote:
> Yep, definitely a bug. Introduced by CASSANDRA-9240 (me; mea culpa).
>
> I've filed a JIRA for you: CASSANDRA-11301
>
> On 3 March 2016 at 14:10, horschi wrote:
>
> > Hi,
> >
> > I just found another on
