You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@guacamole.apache.org by "CHAABANE Foued (Prestataire)" <fo...@externes.justice.gouv.fr> on 2019/07/16 16:03:11 UTC

issue with guacd which using 100% of CPU

Hello,
We have installed guacamole on CentOs7 (4Vcpu, 4Gb mem) 2 years ago.
We access to our servers through guacamole using SSH, RDP and VNC protocols.
Recently we encounter a  problem related to guacd. Guacd processes are not terminated, they continue to run and take a lot of CPU until blocking the guacamole server and disconnect users.

I thing that this problem appears after using many VNC connections. I have not yet found the factor that causes this problem.

Thanks in advance for your help.


Cordialement.

Foued Chaabane
Chef de projet ETD
☎ 01.70.22.74.52/06.20.91.51.98


Re: issue with guacd which using 100% of CPU

Posted by Mike Jumper <mj...@apache.org>.
On Tue, Jul 16, 2019 at 9:03 AM CHAABANE Foued (Prestataire) <
foued.chaabane@externes.justice.gouv.fr> wrote:

> Hello,
>
> We have installed guacamole on CentOs7 (4Vcpu, 4Gb mem) 2 years ago.
>
> We access to our servers through guacamole using SSH, RDP and VNC
> protocols.
>
> Recently we encounter a  problem related to guacd. Guacd processes are not
> terminated, they continue to run and take a lot of CPU until blocking the
> guacamole server and disconnect users.
>

I've seen this occur rarely with RDP connections. If I remember correctly,
the case I saw was due to a bug in FreeRDP where there is an improperly
handled check on a file descriptor, resulting in a tight loop that never
ends. I don't recall which versions of FreeRDP were affected.

On the Guacamole side, guacd should automatically kill the processes of
connections that are disconnected if they don't terminate on their own.
This was added for 1.0.0:

https://issues.apache.org/jira/browse/GUACAMOLE-533

 I thing that this problem appears after using many VNC connections. I have
> not yet found the factor that causes this problem.
>

If this occurs again, the thing to do would be to attach gdb to the
affected process and get a backtrace of all threads:

    (gdb) thread apply all bt

That should help determine where the problem lies and whether it's a bug on
our end.

- Mike