> On Sept. 2, 2016, 2:22 p.m., Jonathan Hurley wrote:
> > ambari-server/src/main/java/org/apache/ambari/server/controller/AmbariCustomCommandExecutionHelper.java,
> >  lines 473-478
> > <https://reviews.apache.org/r/51597/diff/3/?file=1490551#file1490551line473>
> >
> >     I'm curious if we should also try to catch this in the 
> > ServiceComponentHost; don't let it's state move to STARTED if it knows it's 
> > a client.
> 
> Laszlo Puskas wrote:
>     Good point, i am a bit uncertain though that such thing goes into that 
> class. (I'd expect some kind of state machine or other specialized service 
> implementation to decide which are the valid state transitions for 
> hostcomponents.) As this change only affects custom commands i'd leave this 
> as it is; please let me know if this is acceptable.
> 
> Jonathan Hurley wrote:
>     I don't think it should hold up this review as you've fixed the specific 
> issue with the command. I also agree that there should be a state machine to 
> govern this. The problem is that the low budget state machine we currently 
> use doesn't take into account the type of component which the state is being 
> added for. Catching it in the impl will ensure that another mistake can't be 
> made. Having a client in the STARTED state could really mess with other logic 
> in the system.

I am about to register this in a bugdb item.


- Laszlo


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/51597/#review147660
-----------------------------------------------------------


On Sept. 2, 2016, 12:11 p.m., Laszlo Puskas wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/51597/
> -----------------------------------------------------------
> 
> (Updated Sept. 2, 2016, 12:11 p.m.)
> 
> 
> Review request for Ambari, Jonathan Hurley, Sandor Magyari, and Sebastian 
> Toader.
> 
> 
> Bugs: AMBARI-18302
>     https://issues.apache.org/jira/browse/AMBARI-18302
> 
> 
> Repository: ambari
> 
> 
> Description
> -------
> 
> Configuration changes are propagated to host components by issuing a restart 
> of the affected components.
> In case of master components this action will start the affected components 
> (regardless it is in INSTALLED or STARTED state) and the desired state will 
> be set to STARTED.
> In Ambari client components are always expected to be in INSTALLED state. 
> However upon configuration changes, the desired state of these components are 
> also set to STARTED. This may be misleading for ambari API users that 
> determine component states by checking the desired state and actual state. 
> (The actual state in case of client components will always be INSTALLED)
> This issue addresses this problem by not updating the desired state when 
> client components are restarted.
> 
> 
> Diffs
> -----
> 
>   
> ambari-server/src/main/java/org/apache/ambari/server/controller/AmbariCustomCommandExecutionHelper.java
>  8c8ae10 
> 
> Diff: https://reviews.apache.org/r/51597/diff/
> 
> 
> Testing
> -------
> 
> Unit tests running.
> 
> 
> Thanks,
> 
> Laszlo Puskas
> 
>

Reply via email to