mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Alexander Rukletsov" <ruklet...@gmail.com>
Subject Re: Review Request 35702: [WIP] Added /reserve HTTP endpoint to the master.
Date Mon, 22 Jun 2015 13:32:16 GMT

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/35702/#review88781
-----------------------------------------------------------



src/master/http.cpp (lines 519 - 520)
<https://reviews.apache.org/r/35702/#comment141370>

    Some food for thought here. I think the ultimate decision whether to grant a request or
not, should be taken by the allocator. Imagine a funky allocator module that is utterly optimistic
and prefers granting requests and then killing tasks.
    
    With quota (what we have discussed offline) things go even further: a master cannot decide
wether a quota can be fulfilled. We should either expose a lot of stats from allocator so
that master can make a decision or we should change allocator interface so that it allows
querying.



src/master/master.cpp (line 749)
<https://reviews.apache.org/r/35702/#comment141371>

    I think reserve is too abstract and may collide with future actions (think quota). How
about `/dynamic/reserve`?


- Alexander Rukletsov


On June 22, 2015, 5:29 a.m., Michael Park wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/35702/
> -----------------------------------------------------------
> 
> (Updated June 22, 2015, 5:29 a.m.)
> 
> 
> Review request for mesos, Adam B, Benjamin Hindman, Ben Mahler, Jie Yu, Joris Van Remoortere,
and Vinod Kone.
> 
> 
> Bugs: MESOS-2600
>     https://issues.apache.org/jira/browse/MESOS-2600
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> This is still a work in progress, but I'm sharing to gather some feedback around:
> 
> (1) I've added `updateAvailable` to the allocator API. Is this necessary or is there
maybe a better way?
> (2) To determine the amount of available resources, I used: `available = total - (offered
+ used)`.
>     Is this still correct with oversubscription in the picture?
> (3) I'm creating an `Offer.Operation` to perform the necessary updates in the allocator
and such.
>     Feels weird to use an "offer" operation when there's not an actual offer. Is this
fine for now?
> 
> 
> Diffs
> -----
> 
>   include/mesos/master/allocator.hpp 92de1af414321281b00eaa6f129e5e3e2c849448 
>   src/Makefile.am dfebd2b14c9cb45c437509809fdf5ac3b0c8838c 
>   src/master/allocator/mesos/allocator.hpp 6cfa04650d91a80211cfbc0809236f9438926c78 
>   src/master/allocator/mesos/hierarchical.hpp 7097482fc0adad1c177c15c35edd51c10754f89c

>   src/master/http.cpp b893013ddd052cb58c520ac0328f4a5f0fed862e 
>   src/master/master.hpp af83d3e82d2c161b3cc4583e78a8cbbd2f9a4064 
>   src/master/master.cpp 0135c155181546d3cb43e9e05bb874af846d928d 
>   src/master/validation.hpp 469d6f56c3de28a34177124aae81ce24cb4ad160 
>   src/master/validation.cpp 9d128aa1b349b018b8e4a1916434d848761ca051 
>   src/tests/mesos.hpp 9157ac079808d2686592e54ea26a26e6a0825ed3 
>   src/tests/reserve_tests.cpp PRE-CREATION 
> 
> Diff: https://reviews.apache.org/r/35702/diff/
> 
> 
> Testing
> -------
> 
> Added `src/tests/reserve_tests.cpp`.
> 
> 
> Thanks,
> 
> Michael Park
> 
>


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