3 commits

Author	SHA1	Message	Date
Kai Fricke	5a7f6e4fdd	[rfc][ci] create fake docker-compose cluster environment (#20256) ... Following #18987 this PR adds a docker-compose based local multi node cluster. The fake multinode docker comprises two parts. The docker_monitor.py script is a watch script calling docker compose up whenever the docker-compose.yaml changes. The node provider creates and updates the docker compose according to the autoscaling requirements. This mode fully supports autoscaling and comes with test utilities to start and connect to docker-compose autoscaling environments. There's also a sample test case showing how this can be used.	2022-01-11 04:35:36 +00:00
Dmitri Gekhtman	8971422d8f	[autoscaler] Use drain node api in autoscaler before terminating nodes (#20013) ... * wip * Draft * Use bytest for node id * remove stray helm change * fix autoscaler init arg * don't forget to instantiate new load metrics dict * remove extraneous diff * Timeout, comments, function signature. * typo * another comment * tweak * docstring * shorter timeout * Use a better error code * missing self * Dedent example * Add drain node prometheus metric. * comment * Update tests part 1: test_autoscaler.py * Update tests part 2: test_resource_demand_scheduler * lint * Update tests part 3: test_autoscaling_policy * Unit tests for new Prometheus metric and DrainNode error handling. * comment * removed unused function * Try adding ability to mock out process termination to fake node provider * Add integration test. * fix * fix * lint * Improve log message * fix * Simplify test * Fix doc example * remove unused dict * Mock out process termination in a subclass * Add add doc string and comment explaining prune active ips. * Comment: wtf is use_node_id_as_ip * one more comment * more explanation * period * tweak	2021-11-11 08:31:40 -08:00
Eric Liang	9f1cd9e867	[docs] Document fake multi-node autoscaler (#19329)	2021-10-12 15:59:07 -07:00