Running collectors in Docker

Sorry forgot to add host OS Ubuntu Xenial, docker-ce 20.10.3~3-0~ubuntu-xenial.. Container on Centos8 its hit and miss how it happens they have plenty of CPU/RAM/Disk avail

11 hours ago, Michael Baker said:

Container on Centos8

Wait, I may be misunderstanding, but: you've built your own container for this? Why not use our image?

I've run Collectors in my lab almost exclusively in containers and i run into problems with the EA version sometimes. But our image is what is used to monitor the insides of K8s, so while the EA may have issues, the GA version is usually rock solid.

We forked it due to some issues with ancient SDK's and many small bugs + needed to add some other dep's (wmic) in the run time and we got persistent mounts for things like netflow (If you crate /usr/local/logicmonitor/bin/netflow and persitent mount it the script says it already installed and refuses which we have a patch for) which it does not support..

Then there is this:

- Proxy support broken in offical one it does not honour the proxy settting's passed to the container ONLY for the initial python pull but not the install it self

- Netflow ports broken in official one will try bring container up on 2055 we have stuff bound to 2055 that samplicates

- Nbar reverts back to disabled 

- Syslog ports broken same like netflow

- Some odd cert issues with the official one due to DNS caching and IP's changing

Message me if you want to see what we have done?

To explain what it looks like in a bit more detail we have a number of boxes running collectors in docker on the same vlan.. They all run keepalived and samplicator, samplicator is bound to 514 and 2055 on each host which then forwards back to the host IP on 1514 and 2056 which i the ports we customised the agent to use. If we take ear down a collector it will find the existing one in portal download it with default config of 2055/514 and obv breaks

Yeah, i've run into this with our fork of the OTel Collector, the config gets overwritten when the installer runs. If a config is provided through a volume mount, i think it errors out. Definitely worth making it an option to pop in your own config files.