mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Benjamin Bannier <benjamin.bann...@mesosphere.io>
Subject Re: Review Request 69398: Added validation for framework IDs.
Date Wed, 12 Dec 2018 11:29:42 GMT


> On Dec. 11, 2018, 11:10 p.m., Benjamin Mahler wrote:
> > src/master/validation.cpp
> > Line 510 (original), 510 (patched)
> > <https://reviews.apache.org/r/69398/diff/2/?file=2113065#file2113065line510>
> >
> >     add an extra newline?

?


> On Dec. 11, 2018, 11:10 p.m., Benjamin Mahler wrote:
> > src/master/validation.cpp
> > Lines 530-536 (patched)
> > <https://reviews.apache.org/r/69398/diff/2/?file=2113065#file2113065line530>
> >
> >     Do we want to return directly? (not an issue)
> >     
> >     ```
> >     return internal::validateFrameworkId(frameworkInfo);
> >     ```

The format I picked makes the copy & paste to add another validation step trivial; suggest
to keep it this way.


> On Dec. 11, 2018, 11:10 p.m., Benjamin Mahler wrote:
> > src/tests/master_validation_tests.cpp
> > Lines 4686-4688 (patched)
> > <https://reviews.apache.org/r/69398/diff/2/?file=2113066#file2113066line4686>
> >
> >     s/Empty/Unset/?
> >     
> >     Does this let the empty case through? I think we need to deny that when you
consider filesystem path and the sorter path structure.
> >     
> >     Add a test case to ensure empty string is denied?

I added a check for the set but empty case, and we correctly reject such IDs with this patch.

I noticed that in a number of places in the master and scheduler wrapper we maintain handling
for set, but empty IDs which are treated as unset IDs. I believe this is not required anymore
and posted https://reviews.apache.org/r/69557/ to clean this up.


> On Dec. 11, 2018, 11:10 p.m., Benjamin Mahler wrote:
> > src/tests/master_validation_tests.cpp
> > Lines 4690-4692 (patched)
> > <https://reviews.apache.org/r/69398/diff/2/?file=2113066#file2113066line4690>
> >
> >     Is this actually what we do?
> >     
> >     I believe we append to the master ID:
> >     
> >     https://github.com/apache/mesos/blob/1.7.0/src/master/master.cpp#L11531-L11545
> >     
> >     Which is a UUID:
> >     
> >     https://github.com/apache/mesos/blob/1.7.0/src/master/master.cpp#L337-L338
> >     
> >     So today it's `<master uuid> + "-N"`?

I was oversimplifying a little too much here, switching to use a more realistic format now.


- Benjamin


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/69398/#review211210
-----------------------------------------------------------


On Dec. 12, 2018, 12:29 p.m., Benjamin Bannier wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/69398/
> -----------------------------------------------------------
> 
> (Updated Dec. 12, 2018, 12:29 p.m.)
> 
> 
> Review request for mesos, Benjamin Mahler and Meng Zhu.
> 
> 
> Bugs: MESOS-9469
>     https://issues.apache.org/jira/browse/MESOS-9469
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> During scheduler subscription we currently cannot distinguish
> self-assigned from master-assigned framework IDs, so we might allow
> frameworks to subscribe even if the used framework ID was never assigned
> by any master, see MESOS-1719.
> 
> This patch adds some validation of used framework IDs so that they can
> e.g., savely be used in agents when persisting information.
> 
> We also clarify the API to call out that self-assigned framework IDs are
> not explicitly supported.
> 
> 
> Diffs
> -----
> 
>   include/mesos/mesos.proto ef4d785d0bf7b202b4093bad6a478f84c25261f3 
>   include/mesos/v1/mesos.proto 4e8f7a1bb1c47761f0407d2f4094aa24ce8c3fef 
>   src/master/validation.hpp 9af903970795a5c8c479d1984a580e41d91f6c91 
>   src/master/validation.cpp 5fccc9f4fa6a81c08c5013b06c7b7795897a13a5 
>   src/tests/master_validation_tests.cpp 3aa7b2a2dcaf054ad2a67c52462493c97b7178c4 
> 
> 
> Diff: https://reviews.apache.org/r/69398/diff/3/
> 
> 
> Testing
> -------
> 
> `make check`
> 
> 
> Thanks,
> 
> Benjamin Bannier
> 
>


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