mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Michael Park <mp...@apache.org>
Subject Re: Review Request 60062: Prevented reserve/create with hierarchical roles on a non-capable agent.
Date Wed, 14 Jun 2017 21:42:10 GMT


> On June 14, 2017, 9:30 a.m., Neil Conway wrote:
> > A bit more detail in the commit message would be good: for example, how does this
change relate to ca8d33ab12d702612492928c0a5080bd459830bd ?

Added more detail! Thanks.


> On June 14, 2017, 9:30 a.m., Neil Conway wrote:
> > src/master/validation.cpp
> > Lines 1830 (patched)
> > <https://reviews.apache.org/r/60062/diff/3/?file=1750387#file1750387line1830>
> >
> >     Seems unfortunate to have to pass in the `SlaveID` here just for error messages.
I wonder if the callers of `validate` can include the `SlaveID` in the error message they
return when validation fails?

Removed from here and moved up to the caller.


> On June 14, 2017, 9:30 a.m., Neil Conway wrote:
> > src/master/validation.cpp
> > Lines 1856 (patched)
> > <https://reviews.apache.org/r/60062/diff/3/?file=1750387#file1750387line1856>
> >
> >     We don't use single-quotes for capability names elsewhere.

Removed.


> On June 14, 2017, 9:30 a.m., Neil Conway wrote:
> > src/tests/master_validation_tests.cpp
> > Lines 279 (patched)
> > <https://reviews.apache.org/r/60062/diff/3/?file=1750388#file1750388line279>
> >
> >     So we're passing an invalid (unset) `SlaveID` to the `validate` call below?
Seems confusing.

There's no longer an agent id parameter, so this is gone now :)


> On June 14, 2017, 9:30 a.m., Neil Conway wrote:
> > src/tests/master_validation_tests.cpp
> > Lines 331 (patched)
> > <https://reviews.apache.org/r/60062/diff/3/?file=1750388#file1750388line331>
> >
> >     Setting this to `true` suggests to me that the test requires the hierarchical
agent capability, which isn't the case. I wonder if we want something like a `DEFAULT_AGENT_CAPABILITIES`
constant?
> >     
> >     For now I'd just pass in a default-initialized `slave::Capabilities`, without
`hierarchicalRole` set.

Left them all default constructed within the test cases except for the `AgentHierarchicalRoleCapability`
tests.


- Michael


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/60062/#review177911
-----------------------------------------------------------


On June 14, 2017, 2:39 p.m., Michael Park wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/60062/
> -----------------------------------------------------------
> 
> (Updated June 14, 2017, 2:39 p.m.)
> 
> 
> Review request for mesos and Neil Conway.
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> This patch makes it such that reserve / create operations involving
> resources with a hierarchical role are invalid if they are attempted on
> an agent without a HIERARCHICAL_ROLE capability. Such attempts from an
> operator endpoint will result in a `BadRequest` response. On the
> frameworks side, we've opted to not offer resources from
> a non-HIERARCHICAL_ROLE agent to hierarchical roles (ca8d33ab).
> 
> If we were to undo ca8d33ab, an attempt from a framework to perform such
> an operation would be considered invalid due to this patch and the
> operation would be dropped by the master.
> 
> 
> Diffs
> -----
> 
>   src/master/http.cpp 1dcfe6ef00b0e3984deb79a511e665f638661323 
>   src/master/master.cpp 3c6925dc6055fb7cec12ec03bc557ad462863c58 
>   src/master/validation.hpp 6b53e34f67f6a8b338b92db6cb7398f1dccce5a9 
>   src/master/validation.cpp f45f9e816bdaf6304794604570edbd68d5faf715 
>   src/tests/master_validation_tests.cpp 4e7ce74edf0069b9317f869b299694a45e2a62f2 
> 
> 
> Diff: https://reviews.apache.org/r/60062/diff/4/
> 
> 
> Testing
> -------
> 
> Added new tests to `master_validation_tests.cpp` + `make check`
> 
> 
> Thanks,
> 
> Michael Park
> 
>


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