Merge lp:~brian-murray/ubuntu-archive-tools/fix-early-phasing into lp:ubuntu-archive-tools
Status: | Merged |
---|---|
Merged at revision: | 1448 |
Proposed branch: | lp:~brian-murray/ubuntu-archive-tools/fix-early-phasing |
Merge into: | lp:ubuntu-archive-tools |
Diff against target: |
20 lines (+9/-0) 1 file modified
phased-updater (+9/-0) |
To merge this branch: | bzr merge lp:~brian-murray/ubuntu-archive-tools/fix-early-phasing |
Related bugs: |
Reviewer | Review Type | Date Requested | Status |
---|---|---|---|
Steve Langasek | Approve | ||
Review via email: mp+397749@code.launchpad.net |
Commit message
Only start phasing the package after it has been in updates for more than 7 hours.
Description of the change
Its possible for the phased-updater to check for crashes about a version of a package which has "just" been copied to -updates. This can result in the package being immediately set to a phased update percentage of 0% which will not help us gather additional data about the quantity of crashes and is confusing. This has happened a couple of times recently:
https:/
https:/
This change checks the date_published of the source package and does not consider it for phasing if it is less than 7 hours old. Here's an excerpt from a test run:
2021-02-08 16:03:33,192 - INFO - Skipping recently published pkg https:/
2021-02-08 16:03:33,194 - INFO - Skipping recently published pkg https:/
2021-02-08 16:03:33,196 - INFO - Skipping recently published pkg https:/
2021-02-08 16:03:33,197 - INFO - Skipping recently published pkg https:/
2021-02-08 16:03:33,198 - INFO - Skipping recently published pkg https:/
2021-02-08 16:03:33,199 - INFO - Skipping recently published pkg https:/
N.B the times are PST.
The adverse consequence of this is that if a package has a very high crash rate immediately after publishing, we are going to continue pushing the package out to users instead of us failing closed. Why is that a reasonable tradeoff?