perfSONAR User Q&A and Other Discussion

[Andrew Lake <>]

Hi,

Cassandra should also be capping at 1/4 the memory, do you see it going above that amount? I’ve seen cassandra lock-up a few times on some hosts, but don’t know that I’ve collected where the memory was at when that happened. It would be good to know if this is the case. 

Thanks,

Andy

On March 7, 2017 at 6:57:30 AM, Andreas Demetriou () wrote:

Hi Andy,

 

I’m running Centos 7 on an X86_64 with Maddash 2.0 RC3.

For what I can see at the moment as the VM has been upgraded to 26gb the Cassandra seems to be using the most Memory and is getting higher daily. And this is the Java process which seems to be eating away at the memory every day.

 

Thanks,

 

Andreas Demetriou| Placement Network Engineer

 

Gigaclear plc, Windrush Court, Abingdon, Oxfordshire, OX14 1SY

M: 07976649731 T: 01865591142 W: www.gigaclear.com

Gigaclear plc is registered in England and Wales with company number 07476617

 

From: Andrew Lake [mailto:]
Sent: 06 March 2017 16:48
To: Andreas Demetriou <>;
Subject: Re: [perfsonar-user] Maddash server memory leak

 

Hi,

 

I have not seen this, what version are you running? Is it the java MaDDash process that is leaking or are you seeing an explosion of perl processes that it calls? Java should be defaulting to the smaller of 1/4th the physical memory or 1GB as the max, so at worst what you should be seeing is the java process getting stuck when it hits that limit. Just to be sure, because we’ve had stuff like this before, its not a case where you have an i386/i686 host installed with 16GB of memory is it (such hosts can only use 4GB even though you will be able to “see” 16GB installed)?

 

Thanks,

Andy

 

 

 

On March 6, 2017 at 11:26:12 AM, Andreas Demetriou () wrote:

Hi,

 

I have been having some issues with the Maddash server as there seems to be a memory leak.

 

If server has been running for a period of time it seems to eat away at the ram until the server crashes.

RAM capacity at this time is 16gb.

 

Has this behaviour been seen before?

If so is this a bug that has been/will be fixed?

 

 

Kind Regards,

 

Andreas Demetriou| Placement Network Engineer

 

Gigaclear plc, Windrush Court, Abingdon, Oxfordshire, OX14 1SY

M: 07976649731 T: 01865591142 W: www.gigaclear.com

Gigaclear plc is registered in England and Wales with company number 07476617