This release introduces new allocation spec format. After upgrade transparent downgrade is not possible. To downgrade first put Agent in drain mode to remove all allocations.
- Handle invalid resource providers
- Resource provision is fully refactored. From this release all resources use providers as source. Providers are located inside pods and obey pod constraints.
provisionvariables for interpolations- default values in interpolations
/v1/registryAPI
This release reintroduces clustering.
- Clustering
- (API)
GET/v1/cluster/nodes
--idcommand-line option forsoil agent
This release introduces resources.
/v1/status/ping,/v1/agent/reloadand/v1/agent/drainendpoints are available during refactoring.- Clustering is temporary disabled.
- Resources
- Allow to use
pod.{name,namespace}inunitandblobnames. - Allow to use
pod.{name,namespace,target}inunitandblobsources.
- (API) GET /v1/agent/reload -> PUT /v1/agent/reload
- (API) GET /v1/agent/stop -> PUT /v1/agent/stop
- (API) {GET|PUT|DELETE} /v1/registry/pods
- /v1/agent/drain (PUT, DELETE)
- /v1/status/node (GET)
- /v1/status/nodes (GET)
- status -> allocation
- status. = present -> allocation..present = true
- Massive refactorings
- Announce node
- Sync announced nodes
- Use strict drain state
- Remove loud fields from pod status
- Refactor test system
- /v1/status/ping
- /v1/agent/reload
- /v1/agent/stop
- Evaluator runs instructions simultaneously in sequential phases
- Temporary remove Worker pools
Initial release