Re: Review Request 46723: Send error message to the framework when it is not connected.

2016-04-28 Thread Vinod Kone 

---
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/46723/#review130967
---


Ship it!




Ship It!

- Vinod Kone


On April 27, 2016, 11:37 p.m., Anand Mazumdar wrote:
> 
> ---
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/46723/
> ---
> 
> (Updated April 27, 2016, 11:37 p.m.)
> 
> 
> Review request for mesos and Vinod Kone.
> 
> 
> Bugs: MESOS-5181
> https://issues.apache.org/jira/browse/MESOS-5181
> 
> 
> Repository: mesos
> 
> 
> Description
> ---
> 
> This change adds logic to send the framework an error message
> if the master receives a call from the framework after it has
> disconnected/when the master notices a framework disconnection. 
> Since driver based frameworks don't receive heartbeats from the 
> master. There is no way for them to detect this unless a new 
> master is detected that eventually triggers re-registration.
> 
> 
> Diffs
> -
> 
>   src/master/master.cpp ff41da3d077b65b44277e1bbae88c61b7bb88a3d 
> 
> Diff: https://reviews.apache.org/r/46723/diff/
> 
> 
> Testing
> ---
> 
> Added test in r46724
> 
> 
> Thanks,
> 
> Anand Mazumdar
> 
>

Re: Review Request 46723: Send error message to the framework when it is not connected.

2016-04-27 Thread Anand Mazumdar 


> On April 27, 2016, 6:37 p.m., Vinod Kone wrote:
> > src/master/master.cpp, lines 2078-2093
> > 
> >
> > Do we want to do this when the frameork sends a message or right after 
> > we detect a disconnection in `exited()`?  I'm mainly thinking about 
> > scenarios where a framework is not sending any messages but is waiting for 
> > offers indefinitely. I think it should be safe because the scheduler PID is 
> > unique.

Made changes to do it for both cases as the `exited` message might be dropped 
on the floor.


- Anand


---
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/46723/#review130823
---


On April 27, 2016, 11:37 p.m., Anand Mazumdar wrote:
> 
> ---
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/46723/
> ---
> 
> (Updated April 27, 2016, 11:37 p.m.)
> 
> 
> Review request for mesos and Vinod Kone.
> 
> 
> Bugs: MESOS-5181
> https://issues.apache.org/jira/browse/MESOS-5181
> 
> 
> Repository: mesos
> 
> 
> Description
> ---
> 
> This change adds logic to send the framework an error message
> if the master receives a call from the framework after it has
> disconnected/when the master notices a framework disconnection. 
> Since driver based frameworks don't receive heartbeats from the 
> master. There is no way for them to detect this unless a new 
> master is detected that eventually triggers re-registration.
> 
> 
> Diffs
> -
> 
>   src/master/master.cpp ff41da3d077b65b44277e1bbae88c61b7bb88a3d 
> 
> Diff: https://reviews.apache.org/r/46723/diff/
> 
> 
> Testing
> ---
> 
> Added test in r46724
> 
> 
> Thanks,
> 
> Anand Mazumdar
> 
>

Re: Review Request 46723: Send error message to the framework when it is not connected.

2016-04-27 Thread Anand Mazumdar 

---
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/46723/
---

(Updated April 27, 2016, 11:37 p.m.)


Review request for mesos and Vinod Kone.


Changes
---

Review comments from Vinod


Bugs: MESOS-5181
https://issues.apache.org/jira/browse/MESOS-5181


Repository: mesos


Description (updated)
---

This change adds logic to send the framework an error message
if the master receives a call from the framework after it has
disconnected/when the master notices a framework disconnection. 
Since driver based frameworks don't receive heartbeats from the 
master. There is no way for them to detect this unless a new 
master is detected that eventually triggers re-registration.


Diffs (updated)
-

  src/master/master.cpp ff41da3d077b65b44277e1bbae88c61b7bb88a3d 

Diff: https://reviews.apache.org/r/46723/diff/


Testing
---

Added test in r46724


Thanks,

Anand Mazumdar

Re: Review Request 46723: Send error message to the framework when it is not connected.

2016-04-27 Thread Vinod Kone 

---
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/46723/#review130823
---




src/master/master.cpp (lines 2078 - 2093)


Do we want to do this when the frameork sends a message or right after we 
detect a disconnection in `exited()`?  I'm mainly thinking about scenarios 
where a framework is not sending any messages but is waiting for offers 
indefinitely. I think it should be safe because the scheduler PID is unique.



src/master/master.cpp (line 2082)


s/making it abort/causing the scheduler driver to abort/



src/master/master.cpp (line 2084)


s/is not connected/disconnected/


- Vinod Kone


On April 27, 2016, 4:57 a.m., Anand Mazumdar wrote:
> 
> ---
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/46723/
> ---
> 
> (Updated April 27, 2016, 4:57 a.m.)
> 
> 
> Review request for mesos and Vinod Kone.
> 
> 
> Bugs: MESOS-5181
> https://issues.apache.org/jira/browse/MESOS-5181
> 
> 
> Repository: mesos
> 
> 
> Description
> ---
> 
> This change adds logic to send the framework an error message
> if the master receives a call from the framework after it has
> disconnected. Since driver based frameworks don't receive
> heartbeats from the master. There is no way for them to detect
> this unless a new master is detected that eventually triggers
> re-registration.
> 
> 
> Diffs
> -
> 
>   src/master/master.cpp ff41da3d077b65b44277e1bbae88c61b7bb88a3d 
> 
> Diff: https://reviews.apache.org/r/46723/diff/
> 
> 
> Testing
> ---
> 
> Added test in r46724
> 
> 
> Thanks,
> 
> Anand Mazumdar
> 
>

Review Request 46723: Send error message to the framework when it is not connected.

2016-04-26 Thread Anand Mazumdar 

---
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/46723/
---

Review request for mesos and Vinod Kone.


Bugs: MESOS-5181
https://issues.apache.org/jira/browse/MESOS-5181


Repository: mesos


Description
---

This change adds logic to send the framework an error message
if the master receives a call from the framework after it has
disconnected. Since driver based frameworks don't receive
heartbeats from the master. There is no way for them to detect
this unless a new master is detected that eventually triggers
re-registration.


Diffs
-

  src/master/master.cpp ff41da3d077b65b44277e1bbae88c61b7bb88a3d 

Diff: https://reviews.apache.org/r/46723/diff/


Testing
---

Added test in r46724


Thanks,

Anand Mazumdar