updated wallclock regression model

• lifecycle policy for lambda logs updated to 12 months
• improved wallclock regression model, especially for jobs over 1 hour

refresh cache on all prefixes every 5 minutes

in prep for removing the gateway-internal auto-refresh.

job cancel refactor

• improved job cancelling behavior
• no longer a mix of error- and terminated- statuses
• shouldn't turn completed jobs into terminated- status

minor bugfixes

• calcloud no longer errors when it receives an unhandled exit code
• blackboard job duration fixed so it's not -50 years if job has started but not completed
• new memory model version

various minor fixes

• fix for blackboard job completion using wrong time
• blackboard exit reason string improved
• EC2 instance type revisions

minor bugfixes

• remove deprecated job definitions
• add 0-padding to queue names to improve Batch console sorting behavior

move off of optimal ec2 selection

• operations-ready CRDS S3 config
• memory model updates
• major refactor of compute envs and queues to move off of optimal EC2 selection

quickfix for ssl certs in new memory model docker env

• fixes ssl cert issue in dev/test/ops (but not present in sandbox)

better error handling and memory model

• new lambda manually calls refresh cache operations
• enhanced error handling to be synced with CALDP v0.2.0
• several env-specific variables introduced to the terraform
• new cleanup lambda which cleans up s3 for or
• new job prediction model to predict which memory bin a dataset will need to process successfully.

memory ladder and input waiting

• initial version of job definition memory ladder
• when processing placed message, wait for input tar and memory feature control file