0.9.9 on Docker with Reverse Proxy Stable, 0.9.12 on Docker with Reverse Proxy Unstable

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

0.9.9 on Docker with Reverse Proxy Stable, 0.9.12 on Docker with Reverse Proxy Unstable

Jake Staub
Hello,

Environment:
Docker host OS = Ubuntu Server 16.04.2 LTS
Docker version = 17.03.1-ce
Docker compose version = 1.13.0
jwilder/nginx-proxy version = 1.13.0
Reverse proxy settings = jwilder/nginx-proxy default settings
Target connection = Windows 7 RDP
Network environment = LAN (http) & WAN (https)

In the environment above, 0.9.9 performs consistently. Connections are made with the Windows 7 RDP target and are held until intentionally ended.

In the environment above, 0.9.12 (and 0.9.11) performs inconsistently. Connections are made with the Windows 7 RDP target but last only approximately 5 minutes at which point the connection is terminated for the following: "ERROR: User is not responding." The attached screen shot of the guacd log is an example of the log output. Docker-compose scripts that show how each environment is built are also included.

Has anyone else experienced the same behavior? If so, is there a workaround available that produces consistent 0.9.12 performance?

Regards,
Jake Staub

docker-compose.yml_0-9-9 (912 bytes) Download Attachment
docker-compose.yml_latest (896 bytes) Download Attachment
guacd user is not responding.PNG (81K) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: 0.9.9 on Docker with Reverse Proxy Stable, 0.9.12 on Docker with Reverse Proxy Unstable

Jake Staub
Additional environment information:
Hypervisor = ESXi 6.0.0 Update 2
vCPU = 4
Memory = 16 GB

The following link got me interested in performance problems related to JVMs:
https://engineering.linkedin.com/blog/2016/11/application-pauses-when-running-jvm-inside-linux-control-groups

The following image is a composite based on screen shots from the the VMware ESXi web client:
guacamole composite performance

Both versions of Guacamole hammer the network pretty hard when handling video. For some reason 0.9.9 holds the connection while 0.9.12 (& 0.9.11) drops the connection.

It seems there is a compute resource threshold that 0.9.12 crosses which the system as implemented cannot handle. The article above suggests higher CPU quotas may be in order to support JVM stuff.

Regards,
Jake  
Reply | Threaded
Open this post in threaded view
|

Re: 0.9.9 on Docker with Reverse Proxy Stable, 0.9.12 on Docker with Reverse Proxy Unstable

Jake Staub
Troubleshooting steps:
1. Hardware assisted virtualization exposed to guest OS.
2. vCPU increased from 4 to 8.
3. Anchored on 0.9.11 which is the latest version posted on the Docker hub by Glyptodon.

See below for for screenshots of network and CPU loads. Memory loads are not included as there is nothing significant occurring. Increasing vCPUs bumped up the max network load capability for 0.9.11 but the end result was still a disconnect. 0.9.9 network load escalated as well but connection remained solid.

Conclusion: 0.9.11 connection problems are not overwhelmed by an increase in CPU resources.