Re: file descriptor issue with guacd 1.5.5

[Gihan Rajapaksha <gi...@gmail.com>]

if any one got any insights for this i would highly appreciate it. Thanks

On Tue, Jan 27, 2026 at 1:22 PM GihanWork Rajapaksha <gi...@gmail.com>
wrote:

> Hi Guys,
>
> I'm currently investigating an issue where the guacd service becomes
> unresponsive or fails to spawn new connections once it reaches a specific
> resource threshold.
>
> *Operating System:*  Ubuntu 24.04.3 LTS
> *Service Manager:* SysVinit
> *Process Owner:* Root
> *Version:* 1.5.5
> *H/W*: 8 vcpus, 64 GiB memory
>
> *The Issue:* During load testing of a 360 RDP app launches, we observed
> that the parent guacd process (PID 46449) gradually accumulates File
> Descriptors (FDs). In our most recent snapshot, the parent reached *998
> FDs*, while subsequent child processes failed to initialize.
>
> [image: image.png]
>
>
> Upon reaching the system's default soft limit of 1024 FDs, new connection
> attempts result in the following error: ls: cannot access
> '/proc/[PID]/fd': No such file or directory
>
> This suggests that guacd is unable to open the necessary pipes or sockets
> to maintain the session, causing the process to terminate prematurely
> before the system can fully register it. I observed the tomcat logs after
> the FDs ran out and it kept authenticating the frontend user and calling
> guacd but guacd logs shows that it can't handle any more.
>
> On the same server I have nginx and tomcat client installed, this is
> native setup without any customizations. For the tomcat server I have
> given 8GB  heap memory so it has enough room to handle client connections.
> During our testing, the VM remained stable with CPU usage peaking at only
> 10% and memory utilization steady at 18GB
>
> *I have some questions I hope someone can answer or come across *these
> * before. *
>
> *Scaling Best Practices: * Is there a specific ulimit for guacd in
> environments with high concurrency or large clipboard data transfers?
>
> *Process Ownership:* guacd service runs as root. Do you recommend
> migrating to a dedicated non-privileged user for guacd specifically to
> manage these resource limits more granularly?
>
> *Resource Leakage:* Is a parent process holding ~1000 FDs considered
> normal behavior for the main daemon, or does this indicate a potential leak
> in session cleanup?
>
> I would appreciate any insights into recommended configuration changes or
> optimizations that would allow the service to more effectively utilize the
> available hardware capacity while maintaining long-term stability.
> Specifically, I am looking for guidance on tuning guacd to prevent these
> resource bottlenecks in a high-capacity environment
>
> Thanks,
> Regards
> Gihan
>
>
>
> ------------------------------
>
>