Closed kfirufk closed 7 years ago
Can you check if the real application actually sees those? I think this is just an artifact of rkt enter behavior, which is not exactly entering the same namespaces as seen by the running application.
the real application sees those! :) thanks if I need to enter the rkt container to build and test some stuff and I need a proper /proc for that, would I need to open ssh for each container or is there a different way to resolve this?
This probably need some patching on rkt enter side to detect there is a nested mount namespace and to enter it. The tricky part is that this additional mount namespace is optional (systemd creates it) and it must be detected by inspecting the top pid of your app, which is unknown.
As a workaround if you just need /proc and /sys, after rkt-entering you can either mount
them or nsenter
the mount-ns of your app, depending on which binary you have available.
Can you please report this to the rkt bug tracker? It is probably not so easy and widespread to fix it immediately, but it's worth tackling eventually.
of course :) thank you i'm not familiar with nsenter, i'll read about it
rkt github issue: https://github.com/rkt/rkt/issues/3744
Thanks! I've captured my previous comment there, closing this bug then.
Helllo...
I'm using php docker image as the base image for my container, it's in order to add additional extensions to php.
everything works fine.. the image is created.. but when I'm running it and I execute 'rkt enter ID', I notice that /sys, /proc and /sys are not mounted. any ideas why?
this is my build script:
and the systemd file that starts rkt-php-fpm at the following:
[Unit] Description=PHP-FPM Requires=network-online.target local-fs.target After=network-online.target local-fs.target
[Install] WantedBy=multi-user.target
I'm on Container Linux by CoreOS 1465.0.0 (Ladybug) with rkt version 1.27.0 and latest acbuild from git.
thank you! :)