mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Chi Zhang" <chzhc...@gmail.com>
Subject Re: Review Request 37929: Changed provisioner Store API and implementation so it works as a read-through cache.
Date Fri, 04 Sep 2015 00:19:10 GMT


> On Aug. 31, 2015, 9:42 p.m., Chi Zhang wrote:
> > src/slave/containerizer/provisioners/appc/store.hpp, lines 76-81
> > <https://reviews.apache.org/r/37929/diff/1/?file=1059937#file1059937line76>
> >
> >     Maybe not yet the guarantee on ordering since we don't support dependency in
images.
> >     
> >     Mainly I think if we don't have to worry about ordering, it is simpler.
> >     
> >     If backends can't get away from that, we should then be more explicit on what
kind of ordering we provie. BFS, first order traversal or sth else?
> 
> Jiang Yan Xu wrote:
>     This comment is about what the method SHOULD do. And the TODO below says: OK we are
actually not doing it just yet. :)
>     
>     It's important to be clear about what we WILL do in this case because this is the
contract between the provisioner and the store. We need to determine right now where the logic
should go in order to make sure the interface is future proof.
>     
>     The backend is told about the ordering because it itself doesn't have enough information
to otherwise detect it.
>     
>     The ordering is basically "what overwites what in the fact of conflict" so it's not
BFS or how we RESOLVE the dependencies.
> 
> Chi Zhang wrote:
>     I guess I wasn't clear if C, B, D, A and C, D, B, A are both acceptable in your example,
afte reading the comment?
> 
> Jiang Yan Xu wrote:
>     Thanks. You are right I could have been more clear about this. In fact B needs to
go before D so I added the following comment: 
>     (B before D in this example is decided by the order in which A specifies its dependencies)

is it true that the order at which A speicifies dependecies matters?


- Chi


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/37929/#review97157
-----------------------------------------------------------


On Sept. 3, 2015, 12:07 a.m., Jiang Yan Xu wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/37929/
> -----------------------------------------------------------
> 
> (Updated Sept. 3, 2015, 12:07 a.m.)
> 
> 
> Review request for mesos, Chi Zhang, Jie Yu, and Timothy Chen.
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> - i.e., It fetches images transparently when it's not in the local cache.
> - This way, the store doesn't have the sense of "localness" anymore but rather it's an
abstraction that provides access to all discoverable images, no matter where they can be found.
> 
> Some context for motivation of this change can be found at: https://reviews.apache.org/r/37881/
> 
> 
> Diffs
> -----
> 
>   src/slave/containerizer/provisioners/appc/store.hpp e48d91be06410bfc028a7b2ed88218e13adbffee

>   src/slave/containerizer/provisioners/appc/store.cpp fbd1c535d398a4d37c30ba23f5408095c7d35b65

>   src/tests/containerizer/appc_provisioner_tests.cpp 47b66b9c30cefe8f9a8e2c1c1341776c2d235020

> 
> Diff: https://reviews.apache.org/r/37929/diff/
> 
> 
> Testing
> -------
> 
> make check.
> 
> 
> Thanks,
> 
> Jiang Yan Xu
> 
>


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