A single-copy deployment has two leaders
Weik.io 2026.2Two copies of a single-copy deployment led at the same time for 5 seconds or more, and a file or a scheduled job may have been handled twice.
| Kind | Severity | Who sees it |
|---|---|---|
deployment- |
Error while both copies lead, then a warning | Everyone who can read the deployment |
What it means#
Two copies of a deployment that should run one copy at a time led it at the same time. Whatever the leader does may then have run twice, such as a file picked up twice or a scheduled job that ran on two servers.
What raises it#
The leads of two copies of one lock overlapped by 5 seconds or more. It is an error while both still lead. Once the overlap ends, the problem becomes a warning that gives the start and end of the overlap, and it leaves the list 15 minutes after the end. A copy that lost the hub and kept leading stands down when it reconnects.
What to check#
See the leader changes of the last seven days:
$ weikio describe deployment <project>/<integration> --env <environment>Read what the routes did while both led:
$ weikio logs <project>/<integration> --env <environment> --from <start> --to <end>How to fix it#
Check the work the routes did in the overlap, and undo or deduplicate what ran twice. By default a leader that can’t reach the hub keeps leading, which keeps the work going while an agent is cut off. Set whenHubUnreachable: standDown in integration.yaml when doing the work twice is worse than pausing it, and the leader then stands down 15 seconds after the hub last accepted its renewal.