Skip to content

[v25.3.x] datalake/coordinator: add reset topic state escape hatch#29759

Merged
nvartolomei merged 2 commits intoredpanda-data:v25.3.xfrom
vbotbuildovich:backport-pr-29596-v25.3.x-576
Mar 5, 2026
Merged

[v25.3.x] datalake/coordinator: add reset topic state escape hatch#29759
nvartolomei merged 2 commits intoredpanda-data:v25.3.xfrom
vbotbuildovich:backport-pr-29596-v25.3.x-576

Conversation

@vbotbuildovich
Copy link
Copy Markdown
Collaborator

Backport of PR #29596

Operators may need to clear pending files from the datalake coordinator
when Iceberg catalog state becomes inconsistent (e.g. after manual
catalog modifications) which can result in stuck coordinator.

The new CoordinatorResetState RPC is exposed via the admin API with
SUPERUSER authorization and plumbed through the coordinator frontend,
RPC service, and state machine.

(cherry picked from commit c6f5621)
In some cases we want to manually choose from where translation
should resume after a coordinator reset. Add per-partition
last_committed overrides and a reset_all_partitions flag.

By default reset_all_partitions is false and the request is a no-op
unless partition_overrides are provided — only the listed partitions
have their pending entries cleared and last_committed set. Setting
reset_all_partitions to true clears all partitions first, then
applies any overrides. This default-safe design minimizes surprise
and makes it harder to accidentally wipe state.

(cherry picked from commit f44ca9d)
@nvartolomei nvartolomei merged commit 98711f4 into redpanda-data:v25.3.x Mar 5, 2026
31 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

area/build area/redpanda kind/backport PRs targeting a stable branch

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants