On 02/24/2012 03:04 PM, Daniel Stenberg wrote:
> On Thu, 23 Feb 2012, Steven Dake wrote:
>
>> If I run my application for 8 hours, the leak adds up to 300k of
>> memory consumed. I had a look into the code for several hours and
>> don't see an immediate way to fix the problem. I'm not sure even what
>> the problem is, as the channel should free all packets on
>> libssh2_channel_free.
>
> I'm curious if the leak remains if your app exits? I mean, could it be
> packets that are received and are just appended in the incoming package
> queue and then are never discarded from there?
>
If my app exits, libssh2 is no longer being consumed by my application
and the leak is collected by the operating system.
My app uses libssh2, and connects to a sshd daemon on remote virtual
machines.
It sends various commands every 10 seconds to check health of
applications ie:
systemctl status httpd
systemctl start httpd
systemctl stop httpd
The processes purpose is to maintain high availability of the VMs, so in
a best case scenario it would run for hundreds+ of days.
The problem is compounded by the fact that I intend for thousands of
these processes to be run on one machine (so in the case of 1000 VMs,
thats 300k*1000 leaked per 8 hours).
http://www.pacemaker-cloud.org to understand the use case
> 300K during 8 hours seems like a fairly small amount of memory per
> packet (assuming you have traffic semi-often on the connection) which
> would imply that the leak doesn't actually occur that often.
>
yes data per packet seems small but adds up over time
> Let me point out that the ->payload allocated pointed is normally passed
> in to _libssh2_packet_add() where the packet is added to the incoming
> queue and it is subsequently freed when that packet has been handled.
>
Perhaps a packet is coming back that is not handled (and thus not freed).
Regards
-steve
_______________________________________________
libssh2-devel http://cool.haxx.se/cgi-bin/mailman/listinfo/libssh2-devel
Received on 2012-02-25