Some of my dockerized services on an agent host run lots of short-lived containers using the host docker daemon. Every native start of a container triggers a creation of rancher instance which is unnecessary load on rancher server and agent and just a waste of resources.
Is there a possibility not to track some natively started containers, e.g. ignore them in case a particular label is specified?
What do you mean that ever start triggers a creation of “rancher instance”? What is the instance that you are talking about?
Currently, there is no label to exclude syncing your containers on your host within the Rancher server UI.
I mean rancher creates a new record in the database that refers a newly started container. I can also see this container in webui and can retrieve it using API. According to agent logs it also starts receiving some requests which it then process and reply on. If I run many short-lived containers the database table grows quite fast. It’s not a big deal, but still rancher does an unnecessary job in my case.