Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

REX-Ray process disappears unexpectantly or appears as defunct #326

Closed
clintkitson opened this issue Feb 28, 2016 · 4 comments
Closed

REX-Ray process disappears unexpectantly or appears as defunct #326

clintkitson opened this issue Feb 28, 2016 · 4 comments
Milestone

Comments

@clintkitson
Copy link
Member

There may be a situation depending on the Linux distribution where REX-Ray unexpectantly stops running or the process ends up in a defunct state.

In reviewing the current documentation we have, the guidance clearly leans towards using rexray start to start the REX-Ray process for servicing Volume Driver requests. This is actually does not seem like the proper way to run REX-Ray in the long term. This is valid mostly for initially getting the service running. Once the proper configuration is in place, REX-Ray should be stopped and then started properly using the init system.

The docs for installing should also be clear in the manual case. 1) Download 2) move to proper bin dir 3) rexray install 4) init based start

For example on Ubuntu and CoreOS, a systemctl start rexray followed by systemctl status rexray to ensure it started properly.

Simply issuing an interactive CLI command of rexray start is not a consistent approach to daemonizing the REX-Ray process and ensuring it runs through session restarts.

@clintkitson clintkitson added this to the 0.3.2 milestone Feb 28, 2016
@clintkitson
Copy link
Member Author

@akutz What do you think here? 1) Are misleading by specifying start as the command line option? 2) If so should we add some info after this operation to state that it should be ran via an init process? 3) I can work on some doc updates. I don't think I am seeing a section for doing a rexray install after binary download or for starting via init system.

@akutz
Copy link
Member

akutz commented Mar 2, 2016

Hi @clintonskitson,

It's an issue with CoreOS behaving in concert with newer Linux distributions. SystemD is configured to clean up all process when a CGroup exits -- and an SSH session is now part of a CGroup. See this and this for how to change this behavior. Ultimately I think your recommended documentation change as well as a patch to REX-Ray so that rexray start calls SystemD if it's present is the best solution.

akutz added a commit to akutz/rexray that referenced this issue Mar 2, 2016
This patch is a fix for issue rexray#326, where CoreOS is placing SSH sessions
into a CGroup, thus killing any REX-Ray daemons not specifically
launched via SystemD.

This path actually adjusts REX-Ray's behavior for not just CoreOS
systems, but any Linux using SystemD. All SystemD-based systems now
prefer SystemD for service-control-management (SCM) commands.
akutz added a commit to akutz/rexray that referenced this issue Mar 2, 2016
This patch is a fix for issue rexray#326, where CoreOS is placing SSH sessions
into a CGroup, thus killing any REX-Ray daemons not specifically
launched via SystemD.

This path actually adjusts REX-Ray's behavior for not just CoreOS
systems, but any Linux using SystemD. All SystemD-based systems now
prefer SystemD for service-control-management (SCM) commands.
akutz added a commit to akutz/rexray that referenced this issue Mar 2, 2016
This patch is a fix for issue rexray#326, where CoreOS is placing SSH sessions
into a CGroup, thus killing any REX-Ray daemons not specifically
launched via SystemD.

This path actually adjusts REX-Ray's behavior for not just CoreOS
systems, but any Linux using SystemD. All SystemD-based systems now
prefer SystemD for service-control-management (SCM) commands.
akutz added a commit to akutz/rexray that referenced this issue Mar 2, 2016
This patch is a fix for issue rexray#326, where CoreOS is placing SSH sessions
into a CGroup, thus killing any REX-Ray daemons not specifically
launched via SystemD.

This path actually adjusts REX-Ray's behavior for not just CoreOS
systems, but any Linux using SystemD. All SystemD-based systems now
prefer SystemD for service-control-management (SCM) commands.
@akutz
Copy link
Member

akutz commented Mar 2, 2016

Hi @clintonskitson,

I won't close this issue until your documentation PR is merged as well. Then I think we can consider this closed.

@clintkitson
Copy link
Member Author

#327

akutz added a commit to akutz/rexray that referenced this issue Jul 24, 2017
Update Release Branch for 0.3.3 by Merging Master Into It
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

2 participants