mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Andrei Sekretenko <asekrete...@apache.org>
Subject Re: Review Request 72887: Tied the 1st hierarchical allocator process to a process ID "allocator".
Date Mon, 21 Sep 2020 12:52:56 GMT


> On Sept. 18, 2020, 9:41 p.m., Benjamin Mahler wrote:
> > src/master/allocator/mesos/hierarchical.hpp
> > Line 971 (original), 971 (patched)
> > <https://reviews.apache.org/r/72887/diff/1/?file=2239631#file2239631line971>
> >
> >     Let's just go with s/hierarchical_allocator/allocator//
> >     
> >     Note that this means we can only have 1 allocator process in memory, so we wouldn't
be able to have a test with more than 1. I think that's currently the case with the master
process too, so it's probably acceptable.
> >     
> >     As an alternative, we could do:
> >     
> >     ```
> >         : ProcessBase(stripTrailing("(1)", process::ID::generate("allocator"))),
> >     ```
> >     
> >     To make it "allocator", "allocator(2)", "allocator(3)", ...

Stripping `"(1)"` sounds like a better idea. Ajdusted the patch (also went with just "allocator").

I'm almost sure that our tests do not use two allocators at the same time intentionally. 
However, I'm not 100% sure they never let allocator processes slip past the test boundary,
i.e. that all of them wait for the allocator to terminate in the destructors/teardown.


- Andrei


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/72887/#review221895
-----------------------------------------------------------


On Sept. 21, 2020, 2:43 p.m., Andrei Sekretenko wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/72887/
> -----------------------------------------------------------
> 
> (Updated Sept. 21, 2020, 2:43 p.m.)
> 
> 
> Review request for mesos and Benjamin Mahler.
> 
> 
> Bugs: MESOS-10177
>     https://issues.apache.org/jira/browse/MESOS-10177
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> The instances of `HierarchicalAllocatorProcess` launched in the same
> instance of libprocess after that will be assigned IDs like
> "allocator(2)", "allocator(3)" and so on.
> 
> This allows the user to specify a fixed URL like
> `https://localhost:5050/allocator/offer_constraints_debug` when querying
> HTTP endpoints of an allocator in the mesos-master program instead of
> looking up the allocator UPID and using that to compose an URL (like
> `https://localhost:5050/hierarchical_allocator(1)/...`)
> 
> 
> Diffs
> -----
> 
>   src/master/allocator/mesos/hierarchical.hpp 225de160772fe689e07877c895800eb711f211b5

> 
> 
> Diff: https://reviews.apache.org/r/72887/diff/2/
> 
> 
> Testing
> -------
> 
> `make check` on several platforms + tested the debug endpoint manually
> 
> 
> Thanks,
> 
> Andrei Sekretenko
> 
>


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