mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Benjamin Mahler <bmah...@apache.org>
Subject Re: Review Request 68346: Fixed a backoff overflow bug in scheduler authentication retry logic.
Date Thu, 16 Aug 2018 19:24:43 GMT

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/68346/#review207424
-----------------------------------------------------------



Looks good! Just an issue in the future.after, as well as the flag description. We can also
rename the flag here and use a deprecated alias (see below).


src/sched/constants.hpp
Lines 41-42 (original), 41-42 (patched)
<https://reviews.apache.org/r/68346/#comment290805>

    is authenticating



src/sched/flags.hpp
Lines 41-45 (original), 41-45 (patched)
<https://reviews.apache.org/r/68346/#comment290806>

    Looks like in both the agent and scheduler patches, this review should update the approach
description here in the flag help message?



src/sched/flags.hpp
Lines 122-125 (original), 122-125 (patched)
<https://reviews.apache.org/r/68346/#comment290809>

    Either in this review or the next one, we can rename to min and use a deprecated alias
for this old name:
    
    E.g.
    
    https://github.com/apache/mesos/blob/1.6.1/src/slave/flags.cpp#L441



src/sched/sched.cpp
Lines 53 (patched)
<https://reviews.apache.org/r/68346/#comment290807>

    Actually, Future::after doesn't need this include, did you use the free-standing after?



src/sched/sched.cpp
Lines 109 (patched)
<https://reviews.apache.org/r/68346/#comment290808>

    I don't see a use of the free-standing after in your diff?



src/sched/sched.cpp
Lines 425-426 (patched)
<https://reviews.apache.org/r/68346/#comment290810>

    Any reason not to use the same indentation as the agent logic here?



src/sched/sched.cpp
Lines 433-440 (original), 444-458 (patched)
<https://reviews.apache.org/r/68346/#comment290804>

    accessing this->running here without a defer seems unsafe? looks like it will crash
if this gets destructed before the timeout
    
    Probably we can just do a future.discard here, it should be ok to discard the future if
running is false and `_authenticate` will check running.


- Benjamin Mahler


On Aug. 16, 2018, 12:11 a.m., Meng Zhu wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/68346/
> -----------------------------------------------------------
> 
> (Updated Aug. 16, 2018, 12:11 a.m.)
> 
> 
> Review request for mesos, Benjamin Mahler and Gastón Kleiman.
> 
> 
> Bugs: MESOS-9147
>     https://issues.apache.org/jira/browse/MESOS-9147
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> This patch fixed the backoff time calculation
> overflow bug described in MESOS-9147.
> 
> The old approach times out an authentication request after
> `flags.authentication_timeout` and then retries after some
> backoff time. This is not optimal because, if the scheduler
> is going to backoff some time before retry, we might as well
> wait that long for the previous authentication request
> (instead of timeout early).
> 
> This patch combines the authentication timeout and
> authentication retry backoff interval into a single
> wait time interval. Now scheduler will timeout the previous
> authentication request after the wait time interval and
> then immediately retry.
> 
> 
> Diffs
> -----
> 
>   src/sched/constants.hpp 9edb25b38ba8e7e1dbbb4ce4c957bb6bd9f4af81 
>   src/sched/flags.hpp 2492665d44c424ff9f4f73c796520ebc51abbdff 
>   src/sched/sched.cpp 4de76225c73c9c17904512f5a72303d93ec915a7 
> 
> 
> Diff: https://reviews.apache.org/r/68346/diff/3/
> 
> 
> Testing
> -------
> 
> make check
> 
> 
> Thanks,
> 
> Meng Zhu
> 
>


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