> On June 4, 2013, 12:03 a.m., Ben Mahler wrote:
> > src/master/constants.cpp, lines 32-33
> > <https://reviews.apache.org/r/11606/diff/1/?file=300053#file300053line32>
> >
> >     Do you have any data or additional context as to why you're bumping 
> > these?
> >     
> >     30 seconds to respond to a ping seems excessive, perhaps you were just 
> > looking to increase the allowed failures? But even then it would be nice to 
> > hear why. :)
> 
> Brenden Matthews wrote:
>     With EC2, the I/O performance is quite poor and I think 15 seconds might 
> be too small.  I had slaves frequently deactivating just because they were 
> busy.
> 
> Vinod Kone wrote:
>     At Twitter, we have been successfully using 75s (15 * 5) for a long time 
> now. Bumping this up to 300s (30 * 10), means master and scheduler take 4 
> times as long to detect unhealthy slaves. May be the right thing to do is to 
> make these configurable via command line? The only tricky part here is that 
> there is a dependency between this value and ZOOKEEPER_SESSION_TIMEOUT 
> declared in src/detector/detector.hpp.

A command line option would work well. I'll take a look.


- Brenden


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/11606/#review21373
-----------------------------------------------------------


On June 11, 2013, 8:02 p.m., Brenden Matthews wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/11606/
> -----------------------------------------------------------
> 
> (Updated June 11, 2013, 8:02 p.m.)
> 
> 
> Review request for mesos.
> 
> 
> Description
> -------
> 
> Increase slave timeout.
> 
> Review: https://reviews.apache.org/r/11606
> 
> 
> Diffs
> -----
> 
>   src/master/constants.cpp 7bc32eb3bdae717ded63436e47031934384eec23 
>   src/master/master.cpp 60c6d4f88f8024796c5e495d7cc6ddde2b754887 
> 
> Diff: https://reviews.apache.org/r/11606/diff/
> 
> 
> Testing
> -------
> 
> 
> Thanks,
> 
> Brenden Matthews
> 
>

Reply via email to