Support separate PXE rootfs image #528
Conversation
overlay.d/05core/usr/lib/dracut/modules.d/20live/coreos-livepxe-rootfs.sh
Outdated
Show resolved
Hide resolved
overlay.d/05core/usr/lib/dracut/modules.d/20live/coreos-livepxe-rootfs.sh
Outdated
Show resolved
Hide resolved
overlay.d/05core/usr/lib/dracut/modules.d/20live/coreos-livepxe-rootfs.sh
Show resolved
Hide resolved
overlay.d/05core/usr/lib/dracut/modules.d/20live/coreos-livepxe-rootfs.sh
Show resolved
Hide resolved
|
I'm interested to know what's the smallest instance (memory) you can run and have this all work. I wonder if this should put to bed coreos/fedora-coreos-tracker#407. |
bgilbert
changed the title
|
Ready for review. |
overlay.d/05core/usr/lib/dracut/modules.d/20live/coreos-livepxe-rootfs.sh
Outdated
Show resolved
Hide resolved
overlay.d/05core/usr/lib/dracut/modules.d/20live/coreos-livepxe-rootfs.sh
Outdated
Show resolved
Hide resolved
| fi | ||
| # We don't need to verify TLS certificates because we're checking the | ||
| # image hash. | ||
| if ! curl --silent --insecure --location --retry 5 "${rootfs_url}" | \ |
There was a problem hiding this comment.
This is totally fine for now, but down the line would be nice to move the HTTP request to rdcore too. So then stream-hash would become fetch-stage2 --hash-file ... --url ... or something. Gives us better control over the request and error handling. (Honestly piping curl is neat, but it just feels funny to me in general doing it for fundamental things like this :) ).
There was a problem hiding this comment.
This was an experiment in not reinventing wheels, and I feel a little weird about it too. If we decide to move more pieces of the pipeline to rdcore down the road, I'm totally fine with that.
|
By the way, the console logs on a |
Not necessary, but if it wouldn't hurt too much it might make sense to serialize them just to reduce confusion. Interspersed logs often lead to people thinking errors are related to unrelated log messages. |
- If the separate rootfs image was appended as a second initrd, make sure the initramfs and rootfs are from the same build, - else if we were asked to fetch the rootfs over HTTP(S), do so, - else if we're shipping the legacy initramfs image during the deprecation window, add a MOTD, - else fail. If we see the karg to fetch the rootfs, automatically enable network. See coreos/fedora-coreos-tracker#390.
|
Updated! We now correctly handle the case where the user specifies
These days we summarize the logs of failed units after a boot failure, so I'm not too concerned about it. |
This reverts a change in a section heading, which broke an incoming link from coreos/fedora-coreos-config#528.
This reverts a change in a section heading, which broke an incoming link from coreos/fedora-coreos-config#528.
overlay: Add 15rhcos-logrotate (BZ#1780079)
If we see the karg to fetch the rootfs, automatically enable network.
See coreos/fedora-coreos-tracker#390. Requires coreos/coreos-assembler#1608.