Haproxy inside docker container with HTTP mode doesn't work

[tomasz_plonski2]

Hi,

I have few haproxy servers running inside docker containers. All of them expose stats via http so I try to configure dynatrace plugin with http mode using user/password/url. Unfortunatelly there is an error message during plugin initialization, the same from global configuration for all host or with configuration in specified host detail:

Error (('Connection aborted.', FileNotFoundError(2, 'No such file or directory'))) for ....

I found in plugin log information, that problem is correlated with 172 line of plugin code:

File "/opt/dynatrace/oneagent/agent/plugin/plugins/ruxit.python.haproxy/ruxit_haproxy.py", line 172, in initialize

self.url[it] = DockerUtils.replace_port_with_exposed_by_docker(entity.processes[0].properties, url)

Is it anything special in using http mode inside container? I know that socket mode will be introduced in oneagent ver 164 but I suspected that http mode is working.

BR,

Tomek

[maciej_grynda]

Hi Tomek! Could you please share which version of agent/plugin are you using? If you could paste whole stack trace that would be great.

[tomasz_plonski2]

OneAgent version
1.159.309.20190218-095935

I've just noticed that in this environment there is problem with accessing docker API directly from localhost (like plugin works) - Docker API acess is using encrypted TLS connections what is not supported at
this time. I can see in the log many lines with similar error regarding another plugins.

Please find detail in attached log:

ruxitagent-pluginagent-1042801.log

[maciej_grynda · Accepted solution]

I think you should file support issue and provide more info there, like host OS and Docker versions. It might related to docker support, not only to haproxy

[tomasz_plonski2]

Ok, thanks, I will