You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This commit was created on GitHub.com and signed with GitHub’s verified signature.
[v22.4.2] 2026-09-24
Fixed
CSD-299
Removed token regeneration in response to CMR 401 errors. During recent load tests, this behavior, combined with an overly aggressive retry strategy, contributed to overwhelming CMR. This change partially reverts CSD-99, which introduced token regeneration on 401 responses to address an issue where EDL cached invalid tokens. Because the underlying EDL caching issue has since been resolved, token regeneration on 401 responses is no longer necessary.
Added
CSD-299
Added a new CMRCallFailedError error type. This error is thrown for CMR failures after the configured retry attempts, including failures on the first attempt, and should be detectable by StepFunctions for CMR-specific retry behavior.
CUMULUS-5369
The instanceMeta API endpoint now returns icebergAdmins, a list of iceberg admin
usernames read from the ${prefix}-iceberg_admins_list SSM parameter created by the rds-iceberg-replication module. An empty list is returned if the parameter does not exist.
Added an ssm() service and SSM.getParameterValue() helper to @cumulus/aws-client to facilitate
this retrieval.
Changed
CSD-299
Updated CMR retries to explicitly use exponential backoff. This defaults to 0 retries. Retries may be specified using the CMR_RETRIES environment variable. If retries are specified, a backoff factor of 2 and a minTimeout of 5s is used. For 3 retries, this leads to retries at 5, 10 and 20 seconds with total lambda duration of up to 35 seconds of wait.