CageFS remount fixed my issue, but support states that CageFS is not the problem, why?
When you are doing CageFS remount, you think that this is something simple, like removing a fence and building a new one, but the thruth is - it's more complex than that. I'll try to provide an easy explanation for people who doesn't know much about Linux and namespaces and a complex explanation for expirienced Linux users:
Imagine you are in charge of the Racoon City - city where zombies are doing some work for you, thousands of zombies working on hundreds of tiny work-spaces, these work spaces having their own walls so zombies don't escape. Everything was fine, untill one day you notice that zombies on some work-spaces are not doing their job and you can't figure out why. So, you open your teminal, and typing:
and problem is solved.
What you think you did:
Just ruined the walls and build new walls, for some reason that helped, so walls were the problem.
What you actually did:
You just launched a big nuke on the city, killed all zombies, destroyed all walls, whole city was a giant crator. Than you build the city, new zombies appeared and started doing their jobs.
Because everything was done from scratch, you can't say for sure what was the problem.
CageFS remount is a very complex operation, with a few stages, because you can't say which of the stages actually helped, you can't identify the root cause.
1. Cheking usual binaries folders with ldd and adding dependencies to CageFS.
2. Proxyexec service restart.
3. Unmounting everything from current CageFS skeleton.
4. Mouting everything to CageFS skeleton.
5. LVE NameSpaces recreation - all processes inside LVE's are killed. New LVE NameSpaces created with all limits reset to config values.
6. CageFS skeleton and personal directories are mounted into user's LVE NameSpaces.
As you can see, a lot is going on under the hood of CageFS remount, and it's impossible to tell which step helped, so even if CageFS remount helped, it doesn't mean that CageFS was the root cause of your problem.
Please sign in to leave a comment.