Github user aledsage commented on a diff in the pull request:

    https://github.com/apache/brooklyn-server/pull/266#discussion_r71313855
  
    --- Diff: 
software/base/src/main/java/org/apache/brooklyn/entity/software/base/lifecycle/MachineLifecycleEffectorTasks.java
 ---
    @@ -379,20 +418,30 @@ public MachineLocation call() throws Exception {
                 if (!(location instanceof 
LocalhostMachineProvisioningLocation))
                     log.info("Starting {}, obtaining a new location instance 
in {} with ports {}", new Object[]{entity(), location, 
flags.get("inboundPorts")});
                 entity().sensors().set(SoftwareProcess.PROVISIONING_LOCATION, 
location);
    +            Transition expectedState = 
entity().sensors().get(Attributes.SERVICE_STATE_EXPECTED);
    +            
    +            // BROOKLYN-263: see corresponding code in doStop()
    +            if (expectedState != null && (expectedState.getState() == 
Lifecycle.STOPPING || expectedState.getState() == Lifecycle.STOPPED)) {
    +                throw new IllegalStateException("Provisioning aborted 
before even begun for "+entity()+" in "+location+" (presumably by a concurrent 
call to stop");
    +            }
    +            entity().sensors().set(PROVISIONING_TASK_STATE, 
ProvisioningTaskState.RUNNING);
    +            
                 MachineLocation machine;
                 try {
                     machine = Tasks.withBlockingDetails("Provisioning machine 
in " + location, new ObtainLocationTask(location, flags));
    -                if (machine == null)
    -                    throw new NoMachinesAvailableException("Failed to 
obtain machine in " + location.toString());
    -            } catch (Exception e) {
    -                throw Exceptions.propagate(e);
    +                entity().sensors().set(PROVISIONED_MACHINE, machine);
    --- End diff --
    
    I'm willing to take that risk - the description says "internal" etc. The 
problem is that otherwise there is a big gap between when provisioning returns 
and when we add the machine to the location. I worry that start() will not 
always go through all those steps if there has been a concurrent call to stop. 
It might skip executing those subsequent tasks. With the change I made, we set 
DONE and the machine in a finally block so it removes that risk.


---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---

Reply via email to