mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Joris Van Remoortere" <joris.van.remoort...@gmail.com>
Subject Re: Review Request 38110: Quota: Checked sanity of quota set requests.
Date Sun, 15 Nov 2015 15:50:17 GMT

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/38110/#review106560
-----------------------------------------------------------

Ship it!



src/master/master.hpp (line 879)
<https://reviews.apache.org/r/38110/#comment165264>

    How about: 
    ```
    Heuristically tries to determine whether a quota request could reasonabily be satisfied
given the current cluster capacity. 
    
    The goal is to determine whether a user may accidentally request an amount that would
prevent frameworks without a quota from getting any offers. A force flag will allow users
to bypass this check.
    
    The heuristic test whether the total quota, including the new request, does not exceed
the sum of non-static cluster resources.



src/master/master.hpp (line 884)
<https://reviews.apache.org/r/38110/#comment165265>

    s/, what/,as well as which



src/master/master.hpp (line 887)
<https://reviews.apache.org/r/38110/#comment165266>

    s/a bunch of agent nodes may terminate right after/ agents may terminate at any time/



src/master/master.hpp (line 890)
<https://reviews.apache.org/r/38110/#comment165267>

    s/verifies the/verifies that the/



src/master/master.hpp (line 892)
<https://reviews.apache.org/r/38110/#comment165263>

    Let's rename this to `CapacityHeuristic`



src/master/quota_handler.cpp (lines 97 - 101)
<https://reviews.apache.org/r/38110/#comment165273>

    I think we should either re-run the check that the quota for this role doesn't exist,
or in the calculation comment mention that this logic works because master->quotas is guaranteed
not to contain the current role's quota yet.



src/master/quota_handler.cpp (lines 109 - 110)
<https://reviews.apache.org/r/38110/#comment165268>

    Let's add a comment as to what this next section is, as you did with the previous sections.



src/master/quota_handler.cpp (line 110)
<https://reviews.apache.org/r/38110/#comment165269>

    Can we rename this to something like `nonStaticClusterResources`



src/master/quota_handler.cpp (lines 112 - 114)
<https://reviews.apache.org/r/38110/#comment165270>

    Is this really necessary? If so let's add a comment.



src/master/quota_handler.cpp (lines 117 - 118)
<https://reviews.apache.org/r/38110/#comment165272>

    This is a great comment. Can you also make it clear why we actually want to account for
them?



src/master/quota_handler.cpp (line 119)
<https://reviews.apache.org/r/38110/#comment165271>

    Then we can rename this to `nonStaticAgentResources`. It might be so clear we can remove
the first line of the comment above! :-)



src/master/quota_handler.cpp (lines 123 - 126)
<https://reviews.apache.org/r/38110/#comment165274>

    I understand you like the early termination logic as it can reduce the cost of the function
significantly. Can you add a comment here that this is an optimization / early exit of the
final inequality check?



src/master/quota_handler.cpp (line 130)
<https://reviews.apache.org/r/38110/#comment165275>

    hence the request does not pass the heuristic



src/master/quota_handler.cpp (line 131)
<https://reviews.apache.org/r/38110/#comment165276>

    Not enough available cluster capacity to reasonably satisfy quota request. Consider using
the force attribute



src/master/quota_handler.cpp (lines 199 - 200)
<https://reviews.apache.org/r/38110/#comment165277>

    Heuristic capacity check


- Joris Van Remoortere


On Nov. 14, 2015, 3:44 p.m., Alexander Rukletsov wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/38110/
> -----------------------------------------------------------
> 
> (Updated Nov. 14, 2015, 3:44 p.m.)
> 
> 
> Review request for mesos, Bernd Mathiske, Joerg Schad, Joris Van Remoortere, Joseph Wu,
and Michael Park.
> 
> 
> Bugs: MESOS-3074
>     https://issues.apache.org/jira/browse/MESOS-3074
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> Performs a check whether a quota request is reasonable and can be satisfied at the moment.
A precise answer is impossible here, but a sanity check is still helpful, because it allows
us to filter knowingly unsatisfiable requests.
> 
> 
> Diffs
> -----
> 
>   src/master/master.hpp ead8520b7108a0f2c3a0bb11ae7b543897d111a2 
>   src/master/quota_handler.cpp PRE-CREATION 
> 
> Diff: https://reviews.apache.org/r/38110/diff/
> 
> 
> Testing
> -------
> 
> make check (Mac OS X 10.10.4)
> 
> 
> Thanks,
> 
> Alexander Rukletsov
> 
>


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