Dear Nick,

Thank you for your suggestions.
Looking at the server system logs, I found many identical segfault entries:

Nov 28 10:44:55 snc-guacamole kernel: [9829816.814388] guacd[2871114]: segfault 
at 7fdd045f4bf8 ip 00007fd561fe2433 sp 00007fd5045f12e8 error 4 in 
ld-musl-x86_64.so.1[7fd561f9f000+4c000]
Nov 28 10:44:55 snc-guacamole kernel: [9829816.826102] Code: e3 18 ff ff 8b 44 
24 08 48 81 c4 90 00 00 00 5b 5d 41 5c c3 64 48 8b 04 25 00 00 00 00 48 8b 90 
80 00 00 00 89 ff 48 8d 14 fa <48> 39 32 74 07 48 89 32 80 48 42 01 31 c0 c3 48 
85 f6 74 0a b8 16
Nov 28 10:44:56 snc-guacamole kernel: [9829817.741713] guacd[2871125]: segfault 
at 7fdd035fabf8 ip 00007fd561fe2433 sp 00007fd5035f72e8 error 4 in 
ld-musl-x86_64.so.1[7fd561f9f000+4c000]
Nov 28 10:44:56 snc-guacamole kernel: [9829817.748886] Code: e3 18 ff ff 8b 44 
24 08 48 81 c4 90 00 00 00 5b 5d 41 5c c3 64 48 8b 04 25 00 00 00 00 48 8b 90 
80 00 00 00 89 ff 48 8d 14 fa <48> 39 32 74 07 48 89 32 80 48 42 01 31 c0 c3 48 
85 f6 74 0a b8 16

Is it something you are aware of?
Considering that this happens at around 60 connections, could it be in any way 
related to https://www.mail-archive.com/user@guacamole.apache.org/msg12630.html 
?

I cannot reproduce the error currently as this happens only when exceed our 
usual number of users that is far below 60. I will try to deploy next week an 
instance of guacamole with guacd in debug mode and to initiate more than 60 
connections to it.

Best regards,
Paul

From: Nick Couchman <vn...@apache.org>
Sent: Thursday, November 28, 2024 2:50 PM
To: user@guacamole.apache.org
Subject: Re: Maximum number of simultaneous RDP connections

On Thu, Nov 28, 2024 at 8:27 AM HASENOHR Paul 
<paul.hasen...@ec.europa.eu.invalid<mailto:paul.hasen...@ec.europa.eu.invalid>> 
wrote:
Hello,

We have been using Guacamole for the past years with great satisfaction for our 
users. Guacamole is running on Docker (standard images to which a custom CA has 
been added) on a Debian 12 VM.
We noticed today during a training that led to an increase of the number of 
concurrent connections that we could not exceed 55-58 RDP connections. Any 
other attempt to connect to any RDP server (several machines running Ubuntu 
22.04 with xorgxrdp, some of them without any load) returned immediately a 
black screen with the message “You have been disconnected”.

In the guacd logs, any failed connection was logged as follows:
guacd[1]: INFO: Creating new client for protocol "rdp"
guacd[1]: INFO: Connection ID is "$ed73e4c3-ce34-493c-81aa-2fc3810a27fa"
guacd[1916025]: INFO:   Security mode: TLS
guacd[1916025]: INFO:   Resize method: display-update
guacd[1916025]: INFO:   No clipboard line-ending normalization specified. 
Defaulting to preserving the format of all line endings.
guacd[1916025]: INFO:   User "@b2935015-b864-4928-a674-2c26ed1dd45d" joined 
connection "$ed73e4c3-ce34-493c-81aa-2fc3810a27fa" (1 users now present)
guacd[1916025]: INFO:   Recording of session will be saved to 
"/recording//20241128_100346_XXXXXXXXXX-terminal-004-M_N".
guacd[1916025]: INFO:   Loading keymap "base"
guacd[1916025]: INFO:   Loading keymap "en-us-qwerty"
guacd[1]: INFO: Connection "$ed73e4c3-ce34-493c-81aa-2fc3810a27fa" removed.


A handful of things that I can think of to check:
* It's probably worth putting guacd into debug mode ("-L debug" on the command 
line, or add the log configuration to guacd.conf) and see if that yields any 
more useful information on what happens when the connections are failing.
* Since you're running in Docker, check the various Docker limits and see if 
you're running up against something related to container limits - amount of 
memory, number of open files, process, etc. I believe Docker has facilities for 
limiting this similar to the normal "limits" and/or "ulimit" command on Linux, 
but probably slightly different configurations in slightly different places. 
System logs and/or Docker logs might also indicate bumping up against this.
* Also, check your system logs, particularly any facility that might be present 
for core dumps and see if anything is being generated that indicates a segfault 
or other limit.

-Nick

Reply via email to