release: check remote tag before pushing #65555
Conversation
|
This change is |
There was a problem hiding this comment.
Reviewed 2 of 2 files at r1.
Reviewable status:complete! 0 of 0 LGTMs obtained (waiting on @rail)
build/teamcity-common-support.sh, line 35 at r1 (raw file):
git_wrapped() { GIT_SSH_COMMAND="ssh -i .cockroach-teamcity-key" git "$@" }
Why not just make push_to_git more generic?
build/release/teamcity-publish-release.sh, line 62 at r1 (raw file):
tc_start_block "Check remote tag" if git_wrapped ls-remote --exit-code --tags . "${build_name}"; then
Does the . need to be ssh://[email protected]/cockroachdb/cockroach.git so it checks the remote repo and not the local repo?
There was a problem hiding this comment.
Reviewable status:
build/teamcity-common-support.sh, line 35 at r1 (raw file):
Previously, jlinder (James H. Linder) wrote…
git_wrapped() { GIT_SSH_COMMAND="ssh -i .cockroach-teamcity-key" git "$@" }Why not just make
push_to_gitmore generic?
I thought about that, but didn't want to touch multiple invocations. I can easily refactor it for sure.
build/release/teamcity-publish-release.sh, line 62 at r1 (raw file):
Previously, jlinder (James H. Linder) wrote…
Does the
.need to bessh://[email protected]/cockroachdb/cockroach.gitso it checks the remote repo and not the local repo?
. works fine, I double checked it by creating a new tag locally and running it. It exited with code 2, what is expected. I can be more explicit and set the remote.
Previously, prior to pushing the release git tag, we created and pushed most of the release artifacts (docker images, s3). If we run the script multiple times against the same version, it may rewrite the artifacts and will fail pushing the exiting git tag, leaving the release process in a partially released state. This patch adds a check if the release git tag already pushed to the GitHub repo and fails before any artifacts are created. It also refactors the `push_to_git` function to be more generic. Fixes cockroachdb#54814 Release note: None
There was a problem hiding this comment.
Reviewable status:
build/teamcity-common-support.sh, line 35 at r1 (raw file):
Previously, rail (Rail Aliiev) wrote…
I thought about that, but didn't want to touch multiple invocations. I can easily refactor it for sure.
Done.
build/release/teamcity-publish-release.sh, line 62 at r1 (raw file):
Previously, rail (Rail Aliiev) wrote…
.works fine, I double checked it by creating a new tag locally and running it. It exited with code 2, what is expected. I can be more explicit and set the remote.
Done.
There was a problem hiding this comment.
Reviewed 4 of 4 files at r2.
Reviewable status:
build/teamcity-common-support.sh, line 35 at r1 (raw file):
Previously, rail (Rail Aliiev) wrote…
Done.
Cool. I thought that may have been the reason (which I understand).
build/release/teamcity-publish-release.sh, line 62 at r1 (raw file):
Previously, rail (Rail Aliiev) wrote…
Done.
Hrm. When I tested I got these results:
$ git tag -d v21.1.1
Deleted tag 'v21.1.1' (was f6630b3f52)
$ git ls-remote --exit-code --tags . v21.1.1
$ echo $?
2
$ git ls-remote --exit-code --tags ssh://[email protected]/cockroachdb/cockroach.git v21.1.1
f6630b3f524f3a5e52b330220037af01369b1ff2 refs/tags/v21.1.1
$ echo $?
0
Is that what you were expecting?
I also remembered that, since the origin repo on this checkout is from cockroachlabs/release-staging, the origin remote won't have the tag in it when the tag is in cockroachdb/cockroach , so the remote needs to be explicitly stated since git wouldn't know about it otherwise.
There was a problem hiding this comment.
Reviewable status:
build/release/teamcity-publish-release.sh, line 62 at r1 (raw file):
Previously, jlinder (James H. Linder) wrote…
Hrm. When I tested I got these results:
$ git tag -d v21.1.1 Deleted tag 'v21.1.1' (was f6630b3f52) $ git ls-remote --exit-code --tags . v21.1.1 $ echo $? 2 $ git ls-remote --exit-code --tags ssh://[email protected]/cockroachdb/cockroach.git v21.1.1 f6630b3f524f3a5e52b330220037af01369b1ff2 refs/tags/v21.1.1 $ echo $? 0Is that what you were expecting?
I also remembered that, since the origin repo on this checkout is from
cockroachlabs/release-staging, the origin remote won't have the tag in it when the tag is incockroachdb/cockroach, so the remote needs to be explicitly stated since git wouldn't know about it otherwise.
Hmmm, now I don't now how I tested :D
|
bors r+ |
|
Build succeeded: |
In cockroachdb#65555 we added check to make sure we don't try to push the same tag twice. The `git ls-remote` command, which uses `ssh://` was missing a step that configures the SSH key. This patch configures SSH earlier in the process. Release note: None
…65952 #65953 #65958 65397: cloud: add new stanza to specify custom certs for Prometheus r=taroface a=taroface Add a stanza to the Prometheus manifest that allows the user to specify custom certs. The `cockroach-node` specifies the default node secret used by the K8s Operator, as advised by @chrisseto. Relates to cockroachdb/cockroach-operator#469. A doc update to the Prometheus tutorial will refer to this change. 65871: sql: fix a bug which prevented self referencing NOT VALID foreign keys r=fqazi a=ajwerner This bug is due to both using the same object in descs.Txn (sort of) and the fact the fact that MakeMutationComplete does not remove the mutation any longer. I have a feeling, but have not checked, that we lost this back- reference in older versions. Release note (bug fix): Fixed a bug which prevented adding self-referencing FOREIGN KEY constraints in the NOT VALID state. 65938: sql: fix bug in column backfill with virtual NOT NULL columns r=mgartner a=ajwerner Prior to this change we'd inform the column backfiller that it needed to read the virtual columns. These virtual columns don't exist and thus won't be read. If the columns are marked as NOT NULL, then an assertion will fire from inside the row fetcher. This PR fixes the bug by not requesting the virtual columns. Fixes #65915. Release note (bug fix): Fixed a bug which prevented adding columns to tables which contain data and use NOT NULL virtual columns 65943: release: configure ssh key before using git r=rail a=rail In #65555 we added check to make sure we don't try to push the same tag twice. The `git ls-remote` command, which uses `ssh://` was missing a step that configures the SSH key. This patch configures SSH earlier in the process. Release note: None 65945: authors: add JeffSwenson to authors r=JeffSwenson a=JeffSwenson Release note: None 65946: authors: add adwittumuluri to authors r=adwittumuluri a=adwittumuluri Release note: None 65947: authors: add todd to authors r=matthewtodd a=matthewtodd Release note: None 65949: authors: add sarkesian to authors r=AlexTalks a=AlexTalks Release note: None 65950: Adding Toshi to Authors r=noguchitoshi a=noguchitoshi [title] 65952: authors: add "Duoc Nguyen" to authors r=duoclikebook a=duoclikebook Release note: None 65953: add nancy.vargas to authors r=nancy-vargas a=nancy-vargas 65958: authors: add <livlobo> to authors r=livlobo a=livlobo Release note: None Co-authored-by: taroface <[email protected]> Co-authored-by: Andrew Werner <[email protected]> Co-authored-by: Rail Aliiev <[email protected]> Co-authored-by: Jeff Swenson <[email protected]> Co-authored-by: Adwit Tumuluri <[email protected]> Co-authored-by: Matthew Todd <[email protected]> Co-authored-by: Alex Sarkesian <[email protected]> Co-authored-by: Toshi Noguchi <[email protected]> Co-authored-by: Duoc Nguyen <[email protected]> Co-authored-by: Nancy Vargas Balderas <[email protected]> Co-authored-by: Liv Lobo <[email protected]>
In cockroachdb#65555 we added check to make sure we don't try to push the same tag twice. The `git ls-remote` command, which uses `ssh://` was missing a step that configures the SSH key. This patch configures SSH earlier in the process. Release note: None
In cockroachdb#65555 we added check to make sure we don't try to push the same tag twice. The `git ls-remote` command, which uses `ssh://` was missing a step that configures the SSH key. This patch configures SSH earlier in the process. Release note: None
Previously, prior to pushing the release git tag, we created and pushed
most of the release artifacts (docker images, s3). If we run the script
multiple times against the same version, it may rewrite the artifacts
and will fail pushing the exiting git tag, leaving the release process
in a partially released state.
This patch adds a check if the release git tag already pushed to the
GitHub repo and fails before any artifacts are created.
Fixes #54814
Release note: None