mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jiang Yan Xu <...@jxu.me>
Subject Re: Review Request 63831: Fixed a bug that removed the suppressed framework from sorter.
Date Tue, 28 Nov 2017 00:41:53 GMT


> On Nov. 21, 2017, 10:27 a.m., Alexander Rukletsov wrote:
> > src/master/allocator/mesos/hierarchical.cpp
> > Lines 430-437 (original), 431-438 (patched)
> > <https://reviews.apache.org/r/63831/diff/1/?file=1894169#file1894169line431>
> >
> >     Do we need a lambda here?

I kept the original use of lambdas. 

I think the benefit of labmdas here is more about readability: compared to 

```
set<string> removedRoles = oldRoles;
foreach (const string& role, newRoles) {
  result.erase(role);
}
```

The constness and construction of the variable is clearly isolated into a small block in a
long method with many similar variables/code blocks.

However as pointed out by the TODO, we should probably just implement a set difference operator
so all these become one-liners. I'll do it later.


> On Nov. 21, 2017, 10:27 a.m., Alexander Rukletsov wrote:
> > src/tests/scheduler_tests.cpp
> > Lines 1554-1556 (patched)
> > <https://reviews.apache.org/r/63831/diff/1/?file=1894170#file1894170line1554>
> >
> >     How about something like this:
> >     // This test verifies that if a framework decides to suppress its roles on reregisteration,
no offers will be made.

The main difference between `NoOffersOnReregistrationWithAllRolesSuppressed` and `NoOffersWithAllRolesSuppressed`
is the initial condition "a framework initially with no roles suppressing offers". Although
your suggestion implies it, I was emphasizing it. How about:

```
This test verifies that if a framework (initially with no roles suppressed) decides to suppress
offers for its roles on reregisteration, no offers will be made.
```

It's probably simpler to read than my original version.


- Jiang Yan


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/63831/#review191613
-----------------------------------------------------------


On Nov. 15, 2017, 11:05 p.m., Jiang Yan Xu wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/63831/
> -----------------------------------------------------------
> 
> (Updated Nov. 15, 2017, 11:05 p.m.)
> 
> 
> Review request for mesos and Alexander Rukletsov.
> 
> 
> Bugs: MESOS-8223
>     https://issues.apache.org/jira/browse/MESOS-8223
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> In `updateFramework`, we currently treat frameworks suppressing a role the same way as
frameworks moving off a role and this could lead to the framework being removed from the framework
sorter, which is unexpected. We should only remove the framework from a framework sorter when
it's moving away from the corresponding role.
> 
> 
> Diffs
> -----
> 
>   src/master/allocator/mesos/hierarchical.cpp f0f1111a7754e5968288edb15929efc9d244b177

>   src/tests/scheduler_tests.cpp 45fc9c0cfccdb22c2e3e8d5de30c04575814a0e9 
> 
> 
> Diff: https://reviews.apache.org/r/63831/diff/1/
> 
> 
> Testing
> -------
> 
> make check.
> 
> Added a test.
> 
> 
> Thanks,
> 
> Jiang Yan Xu
> 
>


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message