Bamboo 6.6.2 in the official docker container dies together with local agent

[earthquakesan]

Hello!

We are running bamboo in docker in our company with couple of local agents (also in docker). When local agent dies (for example missing dependency or simply when killed), the whole bamboo server goes into restart (running with --restart=always).

When bamboo is terminated all the currently running builds turn grey (i.e. bamboo is losing connection to them).

We are using extended version of official bamboo image with version 6.6.2. Is it known bug?

[Foong]

Local agent is part of Bamboo internal process. It should not be able to die. 

Do you mean that there is a script to kill local agent? If not, how does it die?

[eccenca Buyer]

Hi @Foong!

First of all, I was running bamboo in docker container using docker image provided on your docker hub account. This is an important detail. I had to give up the installation using your image, because of this bug and no comments on the issue. My assumption that nobody is actually using your docker image (correct me if I am wrong).

Bamboo was running 3 local agents. And the problem occured when local agent had to execute a bash script with strict mode (http://redsymbol.net/articles/unofficial-bash-strict-mode/). When something went wrong inside the bash script, the agent execution quits (expected). But additionally what happened is that the whole docker container with bamboo main process died (unexpected).

I could create a test case to reproduce the issue, but I do not see any support for your docker image. Are you using it internally at Atlassian?

[Foong]

My assumption that nobody is actually using your docker image (correct me if I am wrong).

There isn't any stats captured for this but there is 1.5k download shown at https://hub.docker.com/u/atlassian/?page=2

But additionally what happened is that the whole docker container with bamboo main process died (unexpected).

Since the whole docker container died, it looks like the problem is related to the docker too.

Are you using it internally at Atlassian?

No. Atlassian developers are not using it but Support Engineers are using it for testing.

I could create a test case to reproduce the issue

Yes, a test case would be great to understand more about the problem.