v.2.11.0

Improvements:

1. Service dependencies were updated.

Bugs fixed:

1. [LUNA-8929] The key that limits the number of concurrently running deployment jobs per namespace was misspelled as parralel_job_limit in the LAMBDA_DEPLOYMENT_SETTINGS setting and is now named parallel_job_limit. Setting values operators had already saved are renamed by the settings migration, so the limits configured earlier are kept as they are. Config files that set this key in the LAMBDA_DEPLOYMENT_SETTINGS namespace sections have to be updated to the new spelling, see job queue

2. [LUNA-8932] Inactive lambdas cleanup disabled lambdas and jobs that were in use when Prometheus was unavailable: the failed request for the last_activity metric was only logged, and the cleanup then selected lambdas by the stale last_usage values. If Prometheus was down for longer than the interval of the LUNA_LAMBDA_INACTIVE_LAMBDA_CLEAR_INTERVAL (or LUNA_LAMBDA_INACTIVE_JOB_CLEAR_INTERVAL) setting, all non-eternal lambdas (or jobs) were disabled. Now the cleanup is skipped with a warning in the log whenever the last_usage update fails, and it runs as before once Prometheus is available again.

3. [LUNA-8970] Fixed an error where put workflow request failed with the 409 error “Job with name … already exists”.

   Warning

   Rolling back the database migration to the previous version deletes older versions of jobs with the same name: only the job with the latest version is kept for each name.

   Fixed an error where after put workflow request jobs of the previous workflow version returned the new workflow image in the imported_image field.

Migration:

See a migration manual.

[LUNA-8970] See a migration manual.