5.5 KiB
How to Release
Comparing branches or tags
Execute branch.py to compare branches or tags. branch.py has 2 positional arguments:
from_ref the ref to show the path from
to_ref the ref to show the path to
You can find more info by executing branch.py -h
Example:
python branch.py v1.2 master - this line compares the master branch with the release branch v1.2
The output of this command will be as follows:
[- 5e000f4b] release v1.2
[+ c000872d] satellite/payments: coupon value feature
...
[+ ec008dcf] build: Go 1.14.4
- sign means that commit not in the master branch(from_ref)
+ sign means that commit not in v1.2 branch(to_ref)
From the above it follows that commits with a plus sign will be included in the next release v1.3
How to write github changelog
the next step is to create the page on Confluence with our changelog for the release v1.3. Example: Release v1.31 Here we need to post changes for each topic(storj-sim, Uplink, Sattelite, Storage Node, General etc.)
Cutting release branch
Then its time to cut the release branch:
git checkout -b v1.3 - will create and checkout branch v1.3
git push origin v1.3- will push release branch to the repo
Also we need to cut same release branch on tardigrade-satellite-theme repo
git checkout -b v1.3 - will create and checkout branch v1.3
git push origin v1.3- will push release branch to the repo
The next step is to create tag for storj repo using tag-release.sh which is in storj/scripts folder and push it.
Example:
./scripts/tag-release.sh v1.3.0-rc
git push origin v1.3.0-rc
Then verify that the Jenkins job of the build Storj V3 for such tag and branch has finished successfully.
Pay attention to tardigrade-satellite-theme job - it should be successfully finished as well.
How to cherry pick
If you need to cherry-pick something after the release branch has been created then you need to create point release. Make sure that you have the latest changes, checkout the release branch and execute cherry-pick:
git fetch
git checkout -b <xxx>/cherry-pick-v1.xx
git cherry-pick <your commit hash>
You need push and create pull request to the release branch with that commit.
git push origin <xxx>/cherry-pick-v1.xx
After the pull request will be approved, pass all tests and merged you should create new release tag:
./scripts/tag-release.sh v1.3.1
and push the tag to the repo:
git push origin v1.3.1
Verify that the Jenkins job of the build Storj V3 for such tag has finished successfully.
Double check that there was no change on the tardigrade branding in the meantime. Otherwise, the point release might get a broken tardigrade branding. If there are additional commits on the tardigrade branding it is better to create a release branch and revert them. You also have to update the Jenkins job to build from the release branch. Do not forget to change it back to master for the following regular release.
git clone git@github.com:storj/tardigrade-satellite-theme
git reset --hard <your commit hash>
git checkout -b release-v1.3
git push origin release-v1.3
Update Jenkins job.
Revert from release
If revert needed we proceed with next flow: Ask developer to fix problem and push commit to main branch. After that cherry-pick fix to the release branch. Why we do use this flow but not revert from the release branch? It's to prevent situation to fix bug in the main.
Where to find the release binaries
After Jenkins job for this release finished it will automaticaly post this tag on GitHub release page. The status will be Draft.
Update this tag with changelog that you previously created.
For now changelog is generated automatically, but binaries for darwin not. Darwin binaries should be generated manually and added to tag.
Add New Contributors list to the release. To generate it:
git shortlog -sn release-v1.2 | cut -f 2 > ../old.txt && git shortlog -sn release-v1.3 | cut -f 2 > ../new.txt && grep -Fxv -f ../old.txt ../new.txt
Note, to run this command current and previous release should be on your local machine.
Setting the 'Latest' release version
After 100% storagenodes rollout is finished -> new release should be set as 'Latest'.
Which tests do we want to execute
Everything that could break production. From the perspective of a storage node operator the storage node needs to run stable, should not get disqualified or suspended, payout and usage data should be available on the dashboard, graceful exit, garbage collection. Everything that touches on of these topics is most likely worth a test. From the perspective of a satellite operator the durability, availability and accounting is important. This includes the audit and repair system. Also customer and storage node signups should work. Anything in these area is worth a test. For an uplink the network just needs to work. Node selection is critical. If we keep selecting full or bad nodes the uplink will have a hard time to upload enough pieces. It also affects performance.
Forum post changelog
On the forum you need to highlight the most important changes for the storage node operators and describe it a little.