You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@mesos.apache.org by Vinod Kone <vi...@gmail.com> on 2014/10/29 00:08:20 UTC
Review Request 27315: Updated scheduler driver to exponentially
backoff during registration retries.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/27315/
-----------------------------------------------------------
Review request for mesos and Ben Mahler.
Bugs: MESOS-1903
https://issues.apache.org/jira/browse/MESOS-1903
Repository: mesos-git
Description
-------
Uses the same backoff (except no initial backoff) strategy used by the slave during registration.
Diffs
-----
include/mesos/scheduler.hpp 42e4e279d059801cd85955fd04995b60051a2b5e
src/Makefile.am 374f284e1ac839fbcd8a28171b1ff4fbe8a17bd4
src/local/constants.hpp PRE-CREATION
src/local/constants.cpp PRE-CREATION
src/local/flags.hpp 54e88319afc68007ff5d7c0d0179b685ef845c87
src/sched/sched.cpp 0fb8c7bda75545389f8024489b3c76ae115111f4
src/tests/fault_tolerance_tests.cpp a18a41a3e34ff112e04e693447d757403e5013bd
Diff: https://reviews.apache.org/r/27315/diff/
Testing
-------
make check
Thanks,
Vinod Kone
Re: Review Request 27315: Updated scheduler driver to exponentially
backoff during registration retries.
Posted by Mesos ReviewBot <de...@mesos.apache.org>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/27315/#review59106
-----------------------------------------------------------
Patch looks great!
Reviews applied: [27315]
All tests passed.
- Mesos ReviewBot
On Oct. 29, 2014, 11:16 p.m., Vinod Kone wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/27315/
> -----------------------------------------------------------
>
> (Updated Oct. 29, 2014, 11:16 p.m.)
>
>
> Review request for mesos and Ben Mahler.
>
>
> Bugs: MESOS-1903
> https://issues.apache.org/jira/browse/MESOS-1903
>
>
> Repository: mesos-git
>
>
> Description
> -------
>
> Uses the same backoff (except no initial backoff) strategy used by the slave during registration.
>
>
> Diffs
> -----
>
> src/Makefile.am 374f284e1ac839fbcd8a28171b1ff4fbe8a17bd4
> src/sched/constants.hpp PRE-CREATION
> src/sched/constants.cpp PRE-CREATION
> src/sched/flags.hpp PRE-CREATION
> src/sched/sched.cpp 0fb8c7bda75545389f8024489b3c76ae115111f4
> src/tests/fault_tolerance_tests.cpp a18a41a3e34ff112e04e693447d757403e5013bd
>
> Diff: https://reviews.apache.org/r/27315/diff/
>
>
> Testing
> -------
>
> make check
>
>
> Thanks,
>
> Vinod Kone
>
>
Re: Review Request 27315: Updated scheduler driver to exponentially
backoff during registration retries.
Posted by Mesos ReviewBot <de...@mesos.apache.org>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/27315/#review60284
-----------------------------------------------------------
Patch looks great!
Reviews applied: [27315]
All tests passed.
- Mesos ReviewBot
On Nov. 6, 2014, 10:20 p.m., Vinod Kone wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/27315/
> -----------------------------------------------------------
>
> (Updated Nov. 6, 2014, 10:20 p.m.)
>
>
> Review request for mesos and Ben Mahler.
>
>
> Bugs: MESOS-1903
> https://issues.apache.org/jira/browse/MESOS-1903
>
>
> Repository: mesos-git
>
>
> Description
> -------
>
> Uses the same backoff (except no initial backoff) strategy used by the slave during registration.
>
>
> Diffs
> -----
>
> src/Makefile.am 9ab3b9c05d435d18ed1c2966f695857fa205e9fd
> src/sched/constants.hpp PRE-CREATION
> src/sched/constants.cpp PRE-CREATION
> src/sched/flags.hpp PRE-CREATION
> src/sched/sched.cpp e5f828d0bf9dd03a01920634abfae685a7861b44
> src/slave/slave.hpp 5b082fc6956238d27f5d3352a3b40ce778a1a3a4
> src/slave/slave.cpp dbfd1a8101d78dee8ea3ac19d990a6a7892e59be
> src/tests/fault_tolerance_tests.cpp 372c4fdec1ef70991d4d550d8d73ac363436fdb9
>
> Diff: https://reviews.apache.org/r/27315/diff/
>
>
> Testing
> -------
>
> make check
>
>
> Thanks,
>
> Vinod Kone
>
>
Re: Review Request 27315: Updated scheduler driver to exponentially
backoff during registration retries.
Posted by Vinod Kone <vi...@gmail.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/27315/
-----------------------------------------------------------
(Updated Nov. 6, 2014, 10:20 p.m.)
Review request for mesos and Ben Mahler.
Changes
-------
benm's comments. NNFR.
Bugs: MESOS-1903
https://issues.apache.org/jira/browse/MESOS-1903
Repository: mesos-git
Description
-------
Uses the same backoff (except no initial backoff) strategy used by the slave during registration.
Diffs (updated)
-----
src/Makefile.am 9ab3b9c05d435d18ed1c2966f695857fa205e9fd
src/sched/constants.hpp PRE-CREATION
src/sched/constants.cpp PRE-CREATION
src/sched/flags.hpp PRE-CREATION
src/sched/sched.cpp e5f828d0bf9dd03a01920634abfae685a7861b44
src/slave/slave.hpp 5b082fc6956238d27f5d3352a3b40ce778a1a3a4
src/slave/slave.cpp dbfd1a8101d78dee8ea3ac19d990a6a7892e59be
src/tests/fault_tolerance_tests.cpp 372c4fdec1ef70991d4d550d8d73ac363436fdb9
Diff: https://reviews.apache.org/r/27315/diff/
Testing
-------
make check
Thanks,
Vinod Kone
Re: Review Request 27315: Updated scheduler driver to exponentially
backoff during registration retries.
Posted by Vinod Kone <vi...@gmail.com>.
> On Oct. 31, 2014, 10:23 p.m., Ben Mahler wrote:
> > src/sched/constants.hpp, line 18
> > <https://reviews.apache.org/r/27315/diff/2/?file=742036#file742036line18>
> >
> > newline here?
done.
> On Oct. 31, 2014, 10:23 p.m., Ben Mahler wrote:
> > src/sched/sched.cpp, line 252
> > <https://reviews.apache.org/r/27315/diff/2/?file=742039#file742039line252>
> >
> > What is the (s) conveying here?
> >
> > Could you just say, "similar to the slave"?
done.
> On Oct. 31, 2014, 10:23 p.m., Ben Mahler wrote:
> > src/sched/sched.cpp, lines 503-525
> > <https://reviews.apache.org/r/27315/diff/2/?file=742039#file742039line503>
> >
> > How about the following? We can do this for the slave too to make it a bit easier (it's a bit hard to understand currently with 'duration' as a name):
> >
> > ```
> > void doReliableRegistration(Duration maxBackoff)
> > {
> > ...
> >
> > maxBackoff = std::min(maxBackoff, scheduler:: REGISTRATION_RETRY_INTERVAL_DEFAULT_MAX);
> >
> > if (framework.has_failover_timeout()) {
> > ...
> >
> > // Don't approach the failover timeout too closely.
> > maxBackoff = std::min(maxBackoff, timeout.get() / 10);
> > }
> >
> > Duration delay = std::min(
> > duration * ((double) ::random() / RAND_MAX),
> > maxBackoff);
> >
> > VLOG(1) << "will retry registration in " << delay << " if necessary";
> >
> > process::delay(delay, self(), &Self::doReliableRegistration, maxBackoff * 2);
> > }
> > ```
> >
> > Notice that we don't need to store duration_ at all since it's the responsibility of doReliableRegistration to bound the maximum anyway!
> >
> > Could we do this cleanup in the slave too?
done. thanks!
one change though,
Duration delay = maxBackoff * ((double) ::random() / RAND_MAX);
> On Oct. 31, 2014, 10:23 p.m., Ben Mahler wrote:
> > src/sched/sched.cpp, line 506
> > <https://reviews.apache.org/r/27315/diff/2/?file=742039#file742039line506>
> >
> > How about s/duration/timeout/ ?
done.
- Vinod
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/27315/#review59102
-----------------------------------------------------------
On Oct. 29, 2014, 11:16 p.m., Vinod Kone wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/27315/
> -----------------------------------------------------------
>
> (Updated Oct. 29, 2014, 11:16 p.m.)
>
>
> Review request for mesos and Ben Mahler.
>
>
> Bugs: MESOS-1903
> https://issues.apache.org/jira/browse/MESOS-1903
>
>
> Repository: mesos-git
>
>
> Description
> -------
>
> Uses the same backoff (except no initial backoff) strategy used by the slave during registration.
>
>
> Diffs
> -----
>
> src/Makefile.am 374f284e1ac839fbcd8a28171b1ff4fbe8a17bd4
> src/sched/constants.hpp PRE-CREATION
> src/sched/constants.cpp PRE-CREATION
> src/sched/flags.hpp PRE-CREATION
> src/sched/sched.cpp 0fb8c7bda75545389f8024489b3c76ae115111f4
> src/tests/fault_tolerance_tests.cpp a18a41a3e34ff112e04e693447d757403e5013bd
>
> Diff: https://reviews.apache.org/r/27315/diff/
>
>
> Testing
> -------
>
> make check
>
>
> Thanks,
>
> Vinod Kone
>
>
Re: Review Request 27315: Updated scheduler driver to exponentially
backoff during registration retries.
Posted by Ben Mahler <be...@gmail.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/27315/#review59102
-----------------------------------------------------------
Ship it!
Looks good, just once concern about bounding for large failover timeouts and one note about cleanup.
src/sched/constants.hpp
<https://reviews.apache.org/r/27315/#comment100418>
newline here?
src/sched/sched.cpp
<https://reviews.apache.org/r/27315/#comment100428>
What is the (s) conveying here?
Could you just say, "similar to the slave"?
src/sched/sched.cpp
<https://reviews.apache.org/r/27315/#comment100698>
How about the following? We can do this for the slave too to make it a bit easier (it's a bit hard to understand currently with 'duration' as a name):
```
void doReliableRegistration(Duration maxBackoff)
{
...
maxBackoff = std::min(maxBackoff, scheduler:: REGISTRATION_RETRY_INTERVAL_DEFAULT_MAX);
if (framework.has_failover_timeout()) {
...
// Don't approach the failover timeout too closely.
maxBackoff = std::min(maxBackoff, timeout.get() / 10);
}
Duration delay = std::min(
duration * ((double) ::random() / RAND_MAX),
maxBackoff);
VLOG(1) << "will retry registration in " << delay << " if necessary";
process::delay(delay, self(), &Self::doReliableRegistration, maxBackoff * 2);
}
```
Notice that we don't need to store duration_ at all since it's the responsibility of doReliableRegistration to bound the maximum anyway!
Could we do this cleanup in the slave too?
src/sched/sched.cpp
<https://reviews.apache.org/r/27315/#comment100699>
How about s/duration/timeout/ ?
src/sched/sched.cpp
<https://reviews.apache.org/r/27315/#comment100700>
We should bound this as well, to ensure that we don't backoff *too* much!
- Ben Mahler
On Oct. 29, 2014, 11:16 p.m., Vinod Kone wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/27315/
> -----------------------------------------------------------
>
> (Updated Oct. 29, 2014, 11:16 p.m.)
>
>
> Review request for mesos and Ben Mahler.
>
>
> Bugs: MESOS-1903
> https://issues.apache.org/jira/browse/MESOS-1903
>
>
> Repository: mesos-git
>
>
> Description
> -------
>
> Uses the same backoff (except no initial backoff) strategy used by the slave during registration.
>
>
> Diffs
> -----
>
> src/Makefile.am 374f284e1ac839fbcd8a28171b1ff4fbe8a17bd4
> src/sched/constants.hpp PRE-CREATION
> src/sched/constants.cpp PRE-CREATION
> src/sched/flags.hpp PRE-CREATION
> src/sched/sched.cpp 0fb8c7bda75545389f8024489b3c76ae115111f4
> src/tests/fault_tolerance_tests.cpp a18a41a3e34ff112e04e693447d757403e5013bd
>
> Diff: https://reviews.apache.org/r/27315/diff/
>
>
> Testing
> -------
>
> make check
>
>
> Thanks,
>
> Vinod Kone
>
>
Re: Review Request 27315: Updated scheduler driver to exponentially
backoff during registration retries.
Posted by Cody Maloney <co...@mesosphere.io>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/27315/#review59075
-----------------------------------------------------------
LGTM
- Cody Maloney
On Oct. 29, 2014, 11:16 p.m., Vinod Kone wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/27315/
> -----------------------------------------------------------
>
> (Updated Oct. 29, 2014, 11:16 p.m.)
>
>
> Review request for mesos and Ben Mahler.
>
>
> Bugs: MESOS-1903
> https://issues.apache.org/jira/browse/MESOS-1903
>
>
> Repository: mesos-git
>
>
> Description
> -------
>
> Uses the same backoff (except no initial backoff) strategy used by the slave during registration.
>
>
> Diffs
> -----
>
> src/Makefile.am 374f284e1ac839fbcd8a28171b1ff4fbe8a17bd4
> src/sched/constants.hpp PRE-CREATION
> src/sched/constants.cpp PRE-CREATION
> src/sched/flags.hpp PRE-CREATION
> src/sched/sched.cpp 0fb8c7bda75545389f8024489b3c76ae115111f4
> src/tests/fault_tolerance_tests.cpp a18a41a3e34ff112e04e693447d757403e5013bd
>
> Diff: https://reviews.apache.org/r/27315/diff/
>
>
> Testing
> -------
>
> make check
>
>
> Thanks,
>
> Vinod Kone
>
>
Re: Review Request 27315: Updated scheduler driver to exponentially
backoff during registration retries.
Posted by Vinod Kone <vi...@gmail.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/27315/
-----------------------------------------------------------
(Updated Oct. 29, 2014, 11:16 p.m.)
Review request for mesos and Ben Mahler.
Changes
-------
After discussing w/ BenM, decided to move the constants and flags to "src/sched" directory instead of "src/local" to be more precise. Note that the new (event/call) driver in "src/scheduler" doesn't do reliable registration, so there are no updates to it.
Bugs: MESOS-1903
https://issues.apache.org/jira/browse/MESOS-1903
Repository: mesos-git
Description
-------
Uses the same backoff (except no initial backoff) strategy used by the slave during registration.
Diffs (updated)
-----
src/Makefile.am 374f284e1ac839fbcd8a28171b1ff4fbe8a17bd4
src/sched/constants.hpp PRE-CREATION
src/sched/constants.cpp PRE-CREATION
src/sched/flags.hpp PRE-CREATION
src/sched/sched.cpp 0fb8c7bda75545389f8024489b3c76ae115111f4
src/tests/fault_tolerance_tests.cpp a18a41a3e34ff112e04e693447d757403e5013bd
Diff: https://reviews.apache.org/r/27315/diff/
Testing
-------
make check
Thanks,
Vinod Kone
Re: Review Request 27315: Updated scheduler driver to exponentially
backoff during registration retries.
Posted by Vinod Kone <vi...@gmail.com>.
> On Oct. 29, 2014, 7:36 p.m., Dominic Hamon wrote:
> > src/sched/sched.cpp, line 1089
> > <https://reviews.apache.org/r/27315/diff/1/?file=736571#file736571line1089>
> >
> > it looks like you derefence this everywhere.. so why did it need to become a pointer? Why not a concrete instance member variable?
n/a after the restructure.
> On Oct. 29, 2014, 7:36 p.m., Dominic Hamon wrote:
> > src/sched/sched.cpp, line 255
> > <https://reviews.apache.org/r/27315/diff/1/?file=736571#file736571line255>
> >
> > the two places this is referenced it is doubled. Please add a comment why or just store the doubled factor somewhere to simplify the readability of the code.
actually, no need for the double. i just copy pasted from slave code but didn't realize that the semantics are slightly different for driver, i.e., the first retry is tried between [0, b] instead of [0, 2b]. fixed.
- Vinod
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/27315/#review59034
-----------------------------------------------------------
On Oct. 28, 2014, 11:08 p.m., Vinod Kone wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/27315/
> -----------------------------------------------------------
>
> (Updated Oct. 28, 2014, 11:08 p.m.)
>
>
> Review request for mesos and Ben Mahler.
>
>
> Bugs: MESOS-1903
> https://issues.apache.org/jira/browse/MESOS-1903
>
>
> Repository: mesos-git
>
>
> Description
> -------
>
> Uses the same backoff (except no initial backoff) strategy used by the slave during registration.
>
>
> Diffs
> -----
>
> include/mesos/scheduler.hpp 42e4e279d059801cd85955fd04995b60051a2b5e
> src/Makefile.am 374f284e1ac839fbcd8a28171b1ff4fbe8a17bd4
> src/local/constants.hpp PRE-CREATION
> src/local/constants.cpp PRE-CREATION
> src/local/flags.hpp 54e88319afc68007ff5d7c0d0179b685ef845c87
> src/sched/sched.cpp 0fb8c7bda75545389f8024489b3c76ae115111f4
> src/tests/fault_tolerance_tests.cpp a18a41a3e34ff112e04e693447d757403e5013bd
>
> Diff: https://reviews.apache.org/r/27315/diff/
>
>
> Testing
> -------
>
> make check
>
>
> Thanks,
>
> Vinod Kone
>
>
Re: Review Request 27315: Updated scheduler driver to exponentially
backoff during registration retries.
Posted by Dominic Hamon <dh...@twopensource.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/27315/#review59034
-----------------------------------------------------------
src/sched/sched.cpp
<https://reviews.apache.org/r/27315/#comment100315>
the two places this is referenced it is doubled. Please add a comment why or just store the doubled factor somewhere to simplify the readability of the code.
src/sched/sched.cpp
<https://reviews.apache.org/r/27315/#comment100313>
maybe set this on line 503 and then override if necessary. This ensures it is set and is clearer to read.
src/sched/sched.cpp
<https://reviews.apache.org/r/27315/#comment100314>
it looks like you derefence this everywhere.. so why did it need to become a pointer? Why not a concrete instance member variable?
- Dominic Hamon
On Oct. 28, 2014, 4:08 p.m., Vinod Kone wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/27315/
> -----------------------------------------------------------
>
> (Updated Oct. 28, 2014, 4:08 p.m.)
>
>
> Review request for mesos and Ben Mahler.
>
>
> Bugs: MESOS-1903
> https://issues.apache.org/jira/browse/MESOS-1903
>
>
> Repository: mesos-git
>
>
> Description
> -------
>
> Uses the same backoff (except no initial backoff) strategy used by the slave during registration.
>
>
> Diffs
> -----
>
> include/mesos/scheduler.hpp 42e4e279d059801cd85955fd04995b60051a2b5e
> src/Makefile.am 374f284e1ac839fbcd8a28171b1ff4fbe8a17bd4
> src/local/constants.hpp PRE-CREATION
> src/local/constants.cpp PRE-CREATION
> src/local/flags.hpp 54e88319afc68007ff5d7c0d0179b685ef845c87
> src/sched/sched.cpp 0fb8c7bda75545389f8024489b3c76ae115111f4
> src/tests/fault_tolerance_tests.cpp a18a41a3e34ff112e04e693447d757403e5013bd
>
> Diff: https://reviews.apache.org/r/27315/diff/
>
>
> Testing
> -------
>
> make check
>
>
> Thanks,
>
> Vinod Kone
>
>
Re: Review Request 27315: Updated scheduler driver to exponentially
backoff during registration retries.
Posted by Vinod Kone <vi...@gmail.com>.
> On Oct. 29, 2014, 2:43 a.m., Cody Maloney wrote:
> > src/sched/sched.cpp, line 506
> > <https://reviews.apache.org/r/27315/diff/1/?file=736571#file736571line506>
> >
> > maxBackoff is never set if duration.isNone.
good catch! thanks. fixed.
> On Oct. 29, 2014, 2:43 a.m., Cody Maloney wrote:
> > include/mesos/scheduler.hpp, line 443
> > <https://reviews.apache.org/r/27315/diff/1/?file=736566#file736566line443>
> >
> > It would be more precise to use an Optional here rather than write the rules/semantics (if I'm reading the rest of the code correctly) into a pointer.
> >
> > Also means we don't have to manually clean up in the destructor.
i restructured the code and this is n/a.
> On Oct. 29, 2014, 2:43 a.m., Cody Maloney wrote:
> > src/sched/sched.cpp, line 519
> > <https://reviews.apache.org/r/27315/diff/1/?file=736571#file736571line519>
> >
> > It would be nice to use <random> here (The headers at least are around in gcc 4.4).
> >
> > Doing this sort of math creates significantly biased randon numbers.
added a TODO for now. i can fix it here and in slave in a subsequent review.
> On Oct. 29, 2014, 2:43 a.m., Cody Maloney wrote:
> > src/sched/sched.cpp, line 522
> > <https://reviews.apache.org/r/27315/diff/1/?file=736571#file736571line522>
> >
> > nextDuration/durationNext would be a more descriptive variable name.
i'll leave it as is for consistency with the naming in slave.
- Vinod
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/27315/#review58951
-----------------------------------------------------------
On Oct. 28, 2014, 11:08 p.m., Vinod Kone wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/27315/
> -----------------------------------------------------------
>
> (Updated Oct. 28, 2014, 11:08 p.m.)
>
>
> Review request for mesos and Ben Mahler.
>
>
> Bugs: MESOS-1903
> https://issues.apache.org/jira/browse/MESOS-1903
>
>
> Repository: mesos-git
>
>
> Description
> -------
>
> Uses the same backoff (except no initial backoff) strategy used by the slave during registration.
>
>
> Diffs
> -----
>
> include/mesos/scheduler.hpp 42e4e279d059801cd85955fd04995b60051a2b5e
> src/Makefile.am 374f284e1ac839fbcd8a28171b1ff4fbe8a17bd4
> src/local/constants.hpp PRE-CREATION
> src/local/constants.cpp PRE-CREATION
> src/local/flags.hpp 54e88319afc68007ff5d7c0d0179b685ef845c87
> src/sched/sched.cpp 0fb8c7bda75545389f8024489b3c76ae115111f4
> src/tests/fault_tolerance_tests.cpp a18a41a3e34ff112e04e693447d757403e5013bd
>
> Diff: https://reviews.apache.org/r/27315/diff/
>
>
> Testing
> -------
>
> make check
>
>
> Thanks,
>
> Vinod Kone
>
>
Re: Review Request 27315: Updated scheduler driver to exponentially
backoff during registration retries.
Posted by Cody Maloney <co...@mesosphere.io>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/27315/#review58951
-----------------------------------------------------------
include/mesos/scheduler.hpp
<https://reviews.apache.org/r/27315/#comment100173>
It would be more precise to use an Optional here rather than write the rules/semantics (if I'm reading the rest of the code correctly) into a pointer.
Also means we don't have to manually clean up in the destructor.
src/sched/sched.cpp
<https://reviews.apache.org/r/27315/#comment100167>
maxBackoff is never set if duration.isNone.
src/sched/sched.cpp
<https://reviews.apache.org/r/27315/#comment100169>
It would be nice to use <random> here (The headers at least are around in gcc 4.4).
Doing this sort of math creates significantly biased randon numbers.
src/sched/sched.cpp
<https://reviews.apache.org/r/27315/#comment100170>
nextDuration/durationNext would be a more descriptive variable name.
- Cody Maloney
On Oct. 28, 2014, 11:08 p.m., Vinod Kone wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/27315/
> -----------------------------------------------------------
>
> (Updated Oct. 28, 2014, 11:08 p.m.)
>
>
> Review request for mesos and Ben Mahler.
>
>
> Bugs: MESOS-1903
> https://issues.apache.org/jira/browse/MESOS-1903
>
>
> Repository: mesos-git
>
>
> Description
> -------
>
> Uses the same backoff (except no initial backoff) strategy used by the slave during registration.
>
>
> Diffs
> -----
>
> include/mesos/scheduler.hpp 42e4e279d059801cd85955fd04995b60051a2b5e
> src/Makefile.am 374f284e1ac839fbcd8a28171b1ff4fbe8a17bd4
> src/local/constants.hpp PRE-CREATION
> src/local/constants.cpp PRE-CREATION
> src/local/flags.hpp 54e88319afc68007ff5d7c0d0179b685ef845c87
> src/sched/sched.cpp 0fb8c7bda75545389f8024489b3c76ae115111f4
> src/tests/fault_tolerance_tests.cpp a18a41a3e34ff112e04e693447d757403e5013bd
>
> Diff: https://reviews.apache.org/r/27315/diff/
>
>
> Testing
> -------
>
> make check
>
>
> Thanks,
>
> Vinod Kone
>
>