Use latest directory commit timestamp as tag#286
Conversation
Artyop
changed the title
Artyop
force-pushed
the
pr/fix-image-tag-renovate
branch
4 times, most recently
from
38d4a0a to
e79d616
Compare
Artyop
force-pushed
the
pr/fix-image-tag-renovate
branch
from
e79d616 to
d579daa
Compare
|
Bit of an issue with libc version while creating the images it seems |
|
Can you rebase with main branch to pick up #281 to avoid any surprise? |
Artyop
force-pushed
the
pr/fix-image-tag-renovate
branch
from
6bb158e to
b665af2
Compare
|
Rebased from main branch, checking for the key issue on the iptables-20.04-image |
iptables-20.04-image might not be used anymore, as we moved to snapshot build for specific iptables version. |
|
#287 is merged, so you might need to rebase again 😅 , good news is no need to check for key issue 🔑 |
well that sounds like a good deal to me! |
This commit change the way of tagging image, from using git-lfs-tree sha256 to using directory's last commit timestamp. It will permit renovate to sort versions and update properly new images, where renovate wasn't able to do so before due to its inability to sort SHAs. Signed-off-by: Antony Reynaud <[email protected]>
Redhat ubi8:8.9-1136 dnf was generating issue to perform remove of gcc and make. Bumping redhat version to ubi8/ubi (from https://catalog.redhat.com/software/base-images) fixed it. Signed-off-by: Antony Reynaud <[email protected]>
Artyop
force-pushed
the
pr/fix-image-tag-renovate
branch
from
b665af2 to
63ec437
Compare
michi-covalent
left a comment
michi-covalent
left a comment
There was a problem hiding this comment.
it might still be useful to have commit sha as a part of the tag (like $timestamp-$sha or something) so that it's easy to see which commit the image got built from.
This commit adds the short commit after the timestamp. Will help to know which commit the image version will be coming from. Signed-off-by: Antony Reynaud <[email protected]>
|
@michi-covalent added the short commit right after the timestamp separated with a "-", let me know if it sounds good for you |
Following the pr cilium#286, update of the readme about how the tagging mechanic works Signed-off-by: Antony Reynaud <[email protected]>
Following the pr cilium#286, update of the readme about how the tagging mechanic works Signed-off-by: Antony Reynaud <[email protected]>
Following the pr #286, update of the readme about how the tagging mechanic works Signed-off-by: Antony Reynaud <[email protected]>
The image-tools repository switched to a timestamp based versioning scheme for its images with cilium/image-tools#286 in order to allow renovate to start managing the updates of those images. But renovate is not able to properly manage those as it does not recognize their versioning scheme. This PR adds a a renovate package rule for those images with a custom [versioning](https://docs.renovatebot.com/modules/versioning/) config. I tested it in a test repo and renovate managed to detect the dependency and update it: <img width="735" height="186" alt="image" src="https://github.com/user-attachments/assets/b53554df-d851-477d-96e6-a304c67ccff0" /> Signed-off-by: Hadrien Patte <[email protected]>
The image-tools repository switched to a timestamp based versioning scheme for its images with cilium/image-tools#286 in order to allow renovate to start managing the updates of those images. But renovate is not able to properly manage those as it does not recognize their versioning scheme. This PR adds a a renovate package rule for those images with a custom [versioning](https://docs.renovatebot.com/modules/versioning/) config. I tested it in a test repo and renovate managed to detect the dependency and update it: <img width="735" height="186" alt="image" src="https://github.com/user-attachments/assets/b53554df-d851-477d-96e6-a304c67ccff0" /> Signed-off-by: Hadrien Patte <[email protected]>
|
Hi, could this have broken the way that we ensure tags are unique based on the content of the image? This month, after merging b6afb9d and 3e3093f, the existing tag 
|
|
I tracked back to this PR by checking when we changed the scheme for the docker tags, however now I think a bit more it's clear that this couldn't have introduced such a breakage as this change was made over a year ago and we only encountered problems in the last few weeks. I'll search for a newer PR that may have broken the behavior more recently. |
|
👋 The issue we're experiencing is described in more details in #385 (and there's a pending fix for cilium/cilium in cilium/cilium#41710) |
8 participants
This commit change the way of tagging image, from using git-lfs-tree sha256 to using directory's last commit timestamp. It will permit renovate to sort versions and update properly new images, where renovate wasn't able to do so before due to its inability to sort SHAs.
The git log command, when given a specific directory in parameter, retrieve the last commit were changes were made in said directory. This means that we can ensure different image versions same as used to be with git-tree.
Renovate can sort versions from left to right, using a timestamp as a tag value allows it to retrieve the latest version deployed with ease.
Fixes #285