mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Neil Conway <neil.con...@gmail.com>
Subject Re: Review Request 60038: Prevented allocating reservation refinements to non-capable frameworks.
Date Wed, 21 Jun 2017 00:43:02 GMT

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/60038/#review178444
-----------------------------------------------------------



FWIW, I find the commit message a bit confusing: "it's possible to translate the refined reservations
into the old format" -- it seems to me it is explicitly _not_ possible to accurately translate
refined reservations into the old format, which is the reason for this change.

- Neil Conway


On June 19, 2017, 9:52 p.m., Michael Park wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/60038/
> -----------------------------------------------------------
> 
> (Updated June 19, 2017, 9:52 p.m.)
> 
> 
> Review request for mesos, Benjamin Mahler and Neil Conway.
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> When reservation refinements are present, old frameworks without the
> RESERVATION_REFINEMENT capability won't be able to understand the new
> format. While it's possible to translate the refined reservations into
> the old format by "hiding" the intermediate reservations in the "stack",
> this leads to ambiguity when processing RESERVE / UNRESERVE operations.
> This is due to the loss of information when we drop the intermediate
> reservations. Therefore, for now we simply filter out resources with
> refined reservations if the framework does not have the capability.
> 
> 
> Diffs
> -----
> 
>   src/master/allocator/mesos/hierarchical.cpp 775fe4115d5f55be1e8542fa4f94699396d9e2a1

> 
> 
> Diff: https://reviews.apache.org/r/60038/diff/5/
> 
> 
> Testing
> -------
> 
> 
> Thanks,
> 
> Michael Park
> 
>


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