create_disk: Pull ostree-container over 9p, instead of pull-local#2764
create_disk: Pull ostree-container over 9p, instead of pull-local#2764cgwalters merged 1 commit intocoreos:mainfrom
Conversation
This should fix openshift/os#594 Basically I think 9p has a bug where under memory pressure, having it *free* an inode requires allocation, which can fail. This works around that bug by pulling the ostree-container archive instead, which is a single big file. Note that this code path does *not* change the semantics at all for the generated disk. The information about the pulled container is discarded and lost. Actually making use of the container bits natively is the `deploy-via-container` image option which is still experimental, but will be used when we progress the ostree native containers work.
Same approuch of coreos#2764 Signed-off-by: Marcio Xavier <marcioas@br.ibm.com>
| bootfs=$(getconfig "bootfs") | ||
| grub_script=$(getconfig "grub-script") | ||
| ostree=$(getconfig "ostree-repo") | ||
| ostree_container=$(getconfig "ostree-container") |
There was a problem hiding this comment.
Another angle to look at this from is that we are dropping another dependency in the code on an unpacked ostree repository, moving to where "source of truth" is a container image (that has an ostree commit inside still).
|
/retest |
|
@cgwalters: The following test failed, say
Full PR test history. Your PR dashboard. DetailsInstructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository. I understand the commands that are listed here. |
jlebon
left a comment
jlebon
left a comment
There was a problem hiding this comment.
Nice! Tested this locally too (wanted to sanity-check the refs).
2 participants
This should fix openshift/os#594
Basically I think 9p has a bug where under memory pressure, having
it free an inode requires allocation, which can fail.
This works around that bug by pulling the ostree-container archive
instead, which is a single big file.
Note that this code path does not change the semantics at all
for the generated disk. The information about the pulled container
is discarded and lost. Actually making use of the container bits
natively is the
deploy-via-containerimage option which isstill experimental, but will be used when we progress the
ostree native containers work.