The lion's share of your drops are from cross-node timeouts, which require
clock synchronization, so check that first. If your clocks are synced,
that means not only are you showing eager dropping based on time, but
despite the eager dropping you are still facing overload.
That local, non-gc paus
Dikang,
Did you take a look at the heap health on those nodes? A quick heap
histogram or dump would help you figure out if it is related to data
issue(wide rows, or bad model) where few nodes may be coming under heap
pressure and dropping messages.
Thanks,
Roopa
*Regards,*
*Roopa Tangirala*
Hi Dikang,
Do you have any GC logging or metrics you can correlate with the dropped
messages? A 13 second pause sounds like a bad GC pause.
Thanks,
Blake
On January 22, 2017 at 10:37:22 PM, Dikang Gu (dikan...@gmail.com) wrote:
Btw, the C* version is 2.2.5, with several backported patches.
Hi Jaroslav,
That's pretty much expected behavior for the current LWT implementation, which
has problems with key contention (the usage pattern you're describing here).
Typically, you want to avoid having multiple clients doing LWT operations on
the same partition key at the same time.
Thanks,
This vote is being failed for CASSANDRA-13058 (committed after tentative
tag) and CASSANDRA-13025 (patch available).
Vote count was 5 binding +1, 1 binding -1, and one non-binding -1.
I'll re-roll a "Take 5" when CASSANDRA-13025 gets committed, tests
appear stable, and we'll try again.
--
Kind
Indeed I conflated the two - thanks Sylvain.
On Mon, Jan 23, 2017 at 11:19 PM, Sylvain Lebresne wrote:
> On Mon, Jan 23, 2017 at 2:31 AM, Nate McCall wrote:
>
>> What was the resolution on this?
>>
>> Looks like we resolved/Fixed CASSANDRA-13058. Can we re-roll and go again?
>>
>
> As I mentione
On Mon, Jan 23, 2017 at 2:31 AM, Nate McCall wrote:
> What was the resolution on this?
>
> Looks like we resolved/Fixed CASSANDRA-13058. Can we re-roll and go again?
>
As I mentioned, CASSANDRA-13025 is also a regression and should be fix
before we re-roll. It's ready for review if someone's int
