You are viewing a plain text version of this content. The canonical link for it is here.

Posted to user@cassandra.apache.org by Parag Patel <pp...@clearpoolgroup.com> on 2015/10/02 21:54:45 UTC

Off heap memory leak?

We have a 12node Cassandra cluster running on 1.2.12.  Each node is using 1.1TB out of 2TB.  Each node has a min+max heap of 24Gb and the physical server has 48Gb.  Our nodes do not restart during the week, only on the weekend, and we're observing that the off heap memory that is consumed ramps up over the course of the week.  There were a few occasions lately where it consumed so much, it started swapping and OS eventually killed the process (for good reason).

I know we need to upgrade, but I'd like to evaluate if upgrading will fix the problem.  Has anyone experienced this or can anyone provide guidance?

Parag

Re: Off heap memory leak?

Posted by Robert Coli <rc...@eventbrite.com>.

On Fri, Oct 2, 2015 at 12:54 PM, Parag Patel <pp...@clearpoolgroup.com>
wrote:

> We have a 12node Cassandra cluster running on 1.2.12.  Each node is using
> 1.1TB out of 2TB.  Each node has a min+max heap of 24Gb and the physical
> server has 48Gb.  Our nodes do not restart during the week, only on the
> weekend, and we’re observing that the off heap memory that is consumed
> ramps up over the course of the week.  There were a few occasions lately
> where it consumed so much, it started swapping and OS eventually killed the
> process (for good reason).
>

It seems plausible this might be due to an off-heap leak.

As no one is going to fix a leak in 1.2.12, were I you I would upgrade to
the HEAD of 1.2.x and see if the problem goes away.

=Rob