🐛 Setup / as shared mount on alpine #445
Conversation
✅ Deploy Preview for kairos-io canceled.
|
|
Considering the output of "qemu-acceptance-tests (alpine) ", the command "mount --make-rshared /sysroot" failed in the fs stage. I wonder why not use "mount --make-rshared /" when root-partition is under /. Since the command recursively changes the type of all mounts, this should do the trick. Could be the last command before starting k3s... |
right, what I'm trying to be careful here is to not break user-contract, and allow to have that set as soon as possible to avoid any possible race - and user assumption when it can be expected those to be effectively ready in the system. I'd prefer to have those as early as possible (as soon as a fs is set up) but that seems to be tricky. I'll first try to get a baseline and we can revisit this later on. |
mudler
force-pushed
the
shared_path
branch
3 times, most recently
from
44bf26a to
94c91f9
Compare
Good point! |
|
Just for your information: mounts are shared when the command is executed in the network stage (did this for testing purposes in /oem/90_custom.yaml). But this may be very late in the boot process... [partition 2, 4 and 5 are remounted rather late in the boot-process, especially partition 4] |
mudler
force-pushed
the
shared_path
branch
5 times, most recently
from
1d7f1f6 to
e3e5385
Compare
Since this is not handled in the boot process, we do this in our devoted file to setup the rootfs configuration Signed-off-by: Ettore Di Giacinto <mudler@kairos.io>
Signed-off-by: Ettore Di Giacinto <mudler@kairos.io>
mudler
force-pushed
the
shared_path
branch
2 times, most recently
from
8d51f3c to
5177289
Compare
Since this is not handled in the boot process, we do this in our devoted file to setup the rootfs configuration. It also adds an e2e test to verify that the mounts are correctly set
Signed-off-by: Ettore Di Giacinto mudler@kairos.io
Which issue(s) this PR fixes (optional, in
fixes #<issue number>(, fixes #<issue_number>, ...)format, will close the issue(s) when PR gets merged):Fixes #440
Thanks @christianprim for the finding!