- Package and use Slurm 18.08 instead of default from distro repo
- Slurm "master" contains separate controller and login nodes
- Integrate RemoteFS shared file system into Slurm cluster
- Auto feature tagging on Slurm nodes
- Support CentOS 7, Ubuntu 16.04, Ubuntu 18.04 Batch pools as Slurm
node targets
- Unify login and Batch pools on cluster user based on login user
- Auto provision passwordless SSH user on compute nodes with login user
context
- Add slurm cluster commands, including orchestrate command
- Add separate SSH for controller, login, nodes
- Add Slurm configuration doc
- Add Slurm guide
- Add Slurm recipe
- Update usage doc
- Remove deprecated MSI VM extension from monitoring and federation
- Fix pool nodes count on non-existent pool
- Refactor SSH info to allow offsets
- Add fs cluster orchestrate command
- Federation proxy lifecycle management
- Federation lifecycle management
- Federation job submission and management
- Mount Azure File share for auto-rotated log persistence
- FIFO within job support
- Constraint matching
- Federations can be created in "unique job id" mode requiring all
submitted jobs via fed jobs add be unique across the entire federation
- Supports nearly 15K actions per job (in non-unique job id mode)
- Task dependency rewrite engine for federated jobs
- Verify dependencies only within task group
- Uniquely identify task dependencies
- Allow tuning of scheduling behavior options
- Package federation logic on proxy into Docker container
- Full guide/walkthrough for federation feature
- Refactor common code between monitor/fed proxy into resource
- Other doc updates
Fred Park