Create a scheduler/executor implementation for workspace auto on/off.

• Add tableworkspace_lifecycle_job

  CREATETYPEworkspace_lifecycle_operationAS ENUM ('stop','start')CREATETABLEworkspace_lifecycle_job (  id uuidNOT NULL,-- unique job ID  created_attimestamp with time zoneNOT NULL,-- when the job was created  updated_attimestamp with time zoneNOT NULL,-- when the job was last updated  started_attimestamp with time zone,-- when the job was started, null means it's pending  cancelled_attimestamp with time zone,-- when the job was cancelled, null means it hasn't been cancelled  completed_attimestamp with time zone,-- when the job was completed, null means it hasn't finished yet  errortext,-- any errors encountered from the job  workspace_id uuidNOT NULL,-- target workspace  operation workspace_lifecycle_operationNOT NULL,-- start or stop  deadlinetimestamp with time zoneNOT NULL,--  when the job is to be executedPRIMARY KEY (id))

• Add tableworkspace_lifecycle_job_logs

  CREATETABLEworkspace_lifecycle_job_logs (    id uuidNOT NULL,-- unique row id    job_id uuidNOT NULL,-- uuid of the workspace lifecycle job that created it    created_attimestamp with time zoneNOT NULL,-- when the logs were created    level log_levelNOT NULL,-- level of the log    outputcharacter varying(1024)NOT NULL-- content of the log);

• Add required methods toquerier.go:

  • AcquireWorkspaceLifecycleJob acquires the lock for a single workspace lifecycle job that isn't started, completed, or cancelled using SKIP LOCKED.
  • InsertWorkspaceLifecycleJob inserts a new row into theworkspace_lifecycle_job table
  • InsertWorkspaceLifecycleJobLogs inserts a new row into theworkspace_lifecycle_job_logs table
  • UpdateWorkspaceLifecycleJobByID updates a row of theworkspace_lifecycle_job_logs table
  • UpdateWorkspaceLifecycleJobWithCancelByID marks a row of theworkspace_lifecycle_job table as cancelled
  • UpdateWorkspaceLifecycleJobWithCompleteByID marks a row of theworkspace_lifecycle_job table as completed

• Create aWorkspaceLifecycleJobScheduler that schedules starting and stopping workspaces at their scheduled start time/stop time

  • A goroutine tied to the application lifecycle will iterate over all workspaces periodically.
  • If a workspace is running and has a defined auto-stop time, if there is no corresponding auto-stop job pending then create said job.
  • If a workspace is stopped and has a defined auto-start time, if there is no corresponding auto-start job pending then create said job.
  • If an auto-start job is pending for a workspace that has no defined auto-start time, mark the job as cancelled.
  • If an auto-stop job is pending for a workspace that has no defined auto-stop time, mark the job as cancelled.

• Create anWorkspaceLifecycleJobExecutor that executes jobs fromworkspace_lifecycle_job:

  • A goroutine tied to the application lifecycle will continuously wait for a pending workspace lifecycle job that has not been claimed.
  • When an unclaimed job is found:
    • Mark the job as claimed (settingstarted_at andupdated_at)
    • Triggers the workspace stop/start unless the workspace is already in the desired state (stopped for auto-stop, running for auto-start)
    • Mark the job as completed