mina-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Wright, Omari" <Omari.Wri...@solers.com>
Subject RE: [Apache SSHD] Memory leak
Date Fri, 21 Dec 2012 22:06:52 GMT
It looks as if the GC isn't releasing resources. Where should I send this hprof file?

-----Original Message-----
From: Wright, Omari [mailto:Omari.Wright@solers.com] 
Sent: Thursday, December 20, 2012 5:52 PM
To: users@mina.apache.org
Cc: elecharny@apache.org
Subject: RE: [Apache SSHD] Memory leak

The standalone application was run with sftpsubsystem and scpcommand active. At first I didn't
change anything as far as java configuration then I tried adding java opts inside the default
java opts block to enable gc and set the max allocation for ram to 2gb. Tomorrow when I get
in I will work on getting those hprof files.

-----Original Message-----
From: Niklas Gustavsson [mailto:niklas@protocol7.com]
Sent: Thursday, December 20, 2012 4:48 PM
To: users@mina.apache.org
Cc: elecharny@apache.org
Subject: Re: [Apache SSHD] Memory leak

On Thu, Dec 20, 2012 at 10:37 PM, Wright, Omari <Omari.Wright@solers.com>wrote:

> Yes, apache ftpserver and apache sshd are behaving the same way. They 
> both exceed the max heap size set for them until all of the VMs RAM is consumed.
> A ps fauxwww shows the apache sshd process at 3GB Ram while it was 
> only allocated 2. Apache ftpserver is at 5GB even though it was only 
> allocated 3. This maxes out the 8G ram the VM had allocated to it and 
> produces terribly slow interactions with both servers.

Could you possibly profile the JVMs (using for example VisualVM) in order to better track
down what's actually uses up your memory? Do you see any GC runs and are the able to reclaim
any memory? The behavior you're describing is atypical an thus we need more clues into what
could be causing it. I'm assuming your running the standalone servers without any customizations?
You you also provide your configuration and JVM settings?


View raw message