Tags: ananth75/genie
Tags
Improve handling of job remotely marked FAILED Upon detecting the job marked as FAILED by the leader, the agent would shutdown but throw a few errors in the process. This minor change cleans up by skipping the step that tries (and fails) to set the final job status.
Database Changes - Increase date field precision to max 6 decimal points - Remove unused requested_archive_location_prefix field from jobs table - Add archive_status column to jobs table - Add launcher_ext column to jobs table - Add requested_launcher_ext column to jobs table - Add launcher_ext column to commands table - Add attribute converter for JSON columns to get rid of custom code embedded in other converter classes - Unpin H2 version Note: These changes only exist within the data tier and the associated DTOs are not yet plumbed through until we're sure how/what/where we want to do them
Enforce job directory limits for agent jobs * Add service that monitors total max size, max single file size, number of files and kills the job if limits are exceeded. * Introduce new properties allowing server runtime override * Handle job killed by limit exceeded
Decouple gRPC context for file streams from control file stream The gRPC file transfer originates in the handler of a different gRPC stream ("control"). This links the two streams such that if one terminates the other one is also shut down. Decouple the two by creating a new context for the file transfer.
Modify client tests to support changes to the server This is a two stage change due to the docker image needing to be updated first for these tests to pass
Fix handling of non-existent S3 resources The recently introduced support for ranges of S3 resources introduced a regression that would make the REST API return error 500 for a non-existent S3 resource. The correct response code is 404. This commit addresses this by returning a valid S3 resource that returns false if exist() is called.
Bug - Close the S3 object stream The lack of close on the S3 object stream was causing an exhaustion of the S3 client connection pool.
PreviousNext