mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Andrei Sekretenko <asekrete...@mesosphere.io>
Subject Re: Review Request 72019: Updated master::Call::SUBSCRIBE to be served in parallel.
Date Fri, 17 Jan 2020 12:42:18 GMT

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/72019/#review219321
-----------------------------------------------------------




src/master/master.hpp
Lines 1941-1943 (original), 1959-1961 (patched)
<https://reviews.apache.org/r/72019/#comment307470>

    Shouldn't this comment be amended now? Seems to contradict the way `subscribe()` works.



src/master/master.hpp
Lines 1969-1971 (patched)
<https://reviews.apache.org/r/72019/#comment307471>

    Having two public mutable members isn't something that makes reasonoing about the code
simpler:)
     
    Also, having this stuff specific for subscribe will complicate things if we ever need
to add a second readonly request handler with a non-readonly postprocessing.
    
    I'm wondering how difficult it would be to extend the return of `ReadOnlyRequestHandler`
with a postprocessing callback that would be guaranteed to be executed immediately after the
batch in the `Master` context... 
    Might be not worth it right now; in this case, it probably should be added as TODO.


- Andrei Sekretenko


On Jan. 16, 2020, 10:22 p.m., Benjamin Mahler wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/72019/
> -----------------------------------------------------------
> 
> (Updated Jan. 16, 2020, 10:22 p.m.)
> 
> 
> Review request for mesos, Andrei Sekretenko and Greg Mann.
> 
> 
> Bugs: MESOS-9497
>     https://issues.apache.org/jira/browse/MESOS-9497
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> This call is not entirely read-only, unlike the other GET_* v1 master
> calls, and therefore it warranted its own patch.
> 
> The approach used is to have a mutex protected list of new subscribers
> that gets written to during the parallel serving. Afterwards, once
> things are serial again on the Master, we add the new subscribers to
> the masters data structures.
> 
> 
> Diffs
> -----
> 
>   src/common/http.hpp 47a4d6a1ad4897155448a6ba64e789b15a78c7a2 
>   src/master/http.cpp 8a588635e688eb52cd7b8320426dc412e7b44e18 
>   src/master/master.hpp 3074918d677430b588c7765f5ed82f4e324eeff4 
>   src/master/readonly_handler.cpp fbe748d99c2520b520f56afa50dc0b9bd809778d 
> 
> 
> Diff: https://reviews.apache.org/r/72019/diff/1/
> 
> 
> Testing
> -------
> 
> 
> Thanks,
> 
> Benjamin Mahler
> 
>


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