Performance History
Latest Results
Sync and retry submodule updates on every pull
When submodules are enabled, an existing checkout pinned to a commit_sha now runs the submodule step on every pull, even when the commit is already checked out, so an update that failed on an earlier pull is retried. The step runs `git submodule sync --recursive` before `git submodule update --init --recursive` so URL changes in `.gitmodules` are picked up.
digit50:fix/git-storage-pinned-commit-checkout
55 minutes ago
Restore the worker environment after each process pool task
`_run_task_in_subprocess` applied each submit's environment snapshot on top of the worker's existing environment, so a reused worker process kept variables set by earlier tasks. The worker environment is now restored after every task, so each task sees only its own snapshot.
digit50:fix/process-pool-runner-context-per-submit
1 hour ago
Update submodules when checking out a new commit_sha
With `include_submodules=True`, an existing checkout that moves to a new
`commit_sha` only ran `git checkout`, leaving submodule working trees at the
previous pin. It now runs `git submodule update --init --recursive` after the
checkout, matching the `--recurse-submodules` used by the clone and pull paths.
digit50:fix/git-storage-pinned-commit-checkout
3 hours ago
Restore the worker environment after each process pool task
`_run_task_in_subprocess` applied each submit's environment snapshot on top of the worker's existing environment, so a reused worker process kept variables set by earlier tasks. The worker environment is now restored after every task, so each task sees only its own snapshot.
digit50:fix/process-pool-runner-context-per-submit
3 hours ago
Serialize context on every ProcessPoolTaskRunner submit
ProcessPoolTaskRunner cached the serialized context and environment on the
first submit and reused them for the rest of the runner's lifetime, so tags
and settings entered around later submits never reached the subprocess. This
drops the cache and computes both on each submit, as ThreadPoolTaskRunner
effectively does.
Closes #23371
digit50:fix/process-pool-runner-context-per-submit
4 hours ago
Check out a new commit_sha when the git checkout already exists
`GitRepository.is_current_commit` compared `git rev-parse <commit_sha>` with
the configured SHA and never read HEAD, so any full SHA counted as already
checked out and `pull_code` skipped the fetch and checkout. It now resolves
both HEAD and the configured commit in the destination and compares them,
which also lets an abbreviated SHA of the current commit match.
Closes #23370
digit50:fix/git-storage-pinned-commit-checkout
4 hours ago
Fix `prefect deploy` dropping list items templated from step outputs
`apply_values(..., remove_notset=False)` kept unresolved dict values but
silently dropped unresolved list items. `prefect deploy` applies environment
variables to the whole deployment config with `remove_notset=False` before
build steps run, so a list item such as `tags: ["{{ get-commit-hash.stdout }}"]`
was removed before its step output was available. Keep the original list item
when `remove_notset` is False, matching the dict behavior.
MohammadHijjawi97:fix-deploy-list-step-output-placeholders
5 hours ago
Latest Branches
0%
1 hour ago
75e5d3b
digit50:fix/git-storage-pinned-commit-checkout
0%
1 hour ago
8ecdbd2
digit50:fix/process-pool-runner-context-per-submit
-1%
5 hours ago
3d9c6ff
rennf93:feat/fastapi-guard-security-middleware