Skip to content

Allow to kill new container that failed CHECKS#1160

Closed
LTe wants to merge 1 commit intodokku:masterfrom
LTe:kill-new-fix
Closed

Allow to kill new container that failed CHECKS#1160
LTe wants to merge 1 commit intodokku:masterfrom
LTe:kill-new-fix

Conversation

@LTe
Copy link
Copy Markdown
Contributor

@LTe LTe commented May 1, 2015

When the check does not pass there is no reason to close the container
in the correct way. We can do this by using SIGKILL signal.

Related: #1151

When the check does not pass there is no reason to close the container
in the correct way. We can do this by using SIGKILL signal.
@3onyc
Copy link
Copy Markdown
Contributor

3onyc commented May 1, 2015

The checks failing only indicate that the service is not behaving as expected, it may still have things running that need to clean up before exiting.

@LTe
Copy link
Copy Markdown
Contributor Author

LTe commented May 1, 2015

@3onyc So the current implementation is correct?

@3onyc
Copy link
Copy Markdown
Contributor

3onyc commented May 1, 2015

@LTe In my opinion, yes.

@michaelshobbs
Copy link
Copy Markdown
Member

@3onyc This PR restores the previous functionality due to the assumption that the if the new container fails checks it's better to fail quickly as we know something went wrong. Can you provide more information as to why you think the current stop first method is correct?

@michaelshobbs
Copy link
Copy Markdown
Member

@3onyc bump

@3onyc
Copy link
Copy Markdown
Contributor

3onyc commented May 4, 2015

Stop allows a container to properly clean up, CHECKS failing only indicates that something is wrong, and does not indicate the app failed to start completely.

I can't currently think of a specific example, but giving containers a chance to clean up doesn't sound like a bad thing.

@josegonzalez
Copy link
Copy Markdown
Member

I think that's fair. If we have background containers, they may work while the web version won't. Therefore, I'm 👎 on this PR.

@josegonzalez
Copy link
Copy Markdown
Member

@LTe thanks for being patient with us :)

@joshco
Copy link
Copy Markdown
Contributor

joshco commented May 4, 2015

+1 on the decision to keep the stop behavior.

A rails example would be a job processor, such as sideqik. Upon crash, the OSS version of sidekiq can lose jobs that have been loaded into memory. So in a CHECKS failing container, the job processor could have started up and begun loading jobs, only to be forcibly terminated (crash) thus losing those jobs

https://github.com/mperham/sidekiq/wiki/Problems-and-Troubleshooting#my-sidekiq-process-is-crashing-what-do-i-do

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants