<#7371 fix(executor): wire previous checkpoint pat...
# flyte-github
a
#7371 fix(executor): wire previous checkpoint path for in-place retries Pull request opened by AdilFayyaz Why are the changes needed? On retries (
Status.Attempts >= 2
),
NewTaskExecutionContext
passed "" as the previous checkpoint path to
NewCheckpointRemoteFilePaths
. As a result,
{{.PrevCheckpointPrefix}}
substituted to empty in pod args,
checkpoint.load_sync()
always returned
None
, and tasks restarted from scratch on every retry. Mirrors the worker-side fix from unionai/cloud#15143. What changes were proposed in this pull request? When
Status.Attempts >= 2
, derive the previous checkpoint path via
ComputeActionOutputPath(..., attempt-1) + ConstructCheckpointPath
and pass it through. How was this patch tested? •
go test ./executor/pkg/plugin/... ./executor/pkg/controller/...
• End-to-end sdk example:
examples/checkpoint/generic_data_checkpoint.py
succeeds on attempt 4 Setup process Screenshots Check all the applicable boxes • I updated the documentation accordingly. • All new and existing tests passed. • All commits are signed-off. Related PRs Stack If you do use
git town
to manage PR Stacks, the stack relevant to this PR will show below. Otherwise, you can ignore this section. Docs link flyteorg/flyte