> On Jan. 6, 2017, 11:18 p.m., Avinash sridharan wrote:
> > src/slave/containerizer/mesos/provisioner/docker/store.cpp, line 268
> > <https://reviews.apache.org/r/54215/diff/4/?file=1596399#file1596399line268>
> >
> >     Given that the `layerPaths` are meant to be unique why not make the 
> > `layerPaths` itself a `hashset`. Will be more explicit in terms of its 
> > usage.

We need to keep the layer ids in order. There is dependency on that.


- Gilbert


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/54215/#review160808
-----------------------------------------------------------


On Jan. 3, 2017, 2:47 p.m., Gilbert Song wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/54215/
> -----------------------------------------------------------
> 
> (Updated Jan. 3, 2017, 2:47 p.m.)
> 
> 
> Review request for mesos, Avinash sridharan, Artem Harutyunyan, Jie Yu, Qian 
> Zhang, and Zhitao Li.
> 
> 
> Bugs: MESOS-6654
>     https://issues.apache.org/jira/browse/MESOS-6654
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> This issue is exposed by pulling the 'mesosphere/inky' docker
> image using registry puller. Due to the duplicate layer id
> from the manifest, there are duplicate layer pathes passed
> to the backend. The aufs backend cannot handle this case and
> returns 'invalid arguments' error. Ideally, we should make
> sure that layer paths that are passed to the backend are
> unique.
> 
> 
> Diffs
> -----
> 
>   src/slave/containerizer/mesos/provisioner/docker/store.cpp 
> 9dccd0673dbc0c61abfd4b88097f86e7d7131c46 
> 
> Diff: https://reviews.apache.org/r/54215/diff/
> 
> 
> Testing
> -------
> 
> make check
> 
> Tested by the unit test 
> `ROOT_CURL_INTERNET_DockerDefaultEntryptRegistryPuller`.
> 
> Manually tested using the `mesosphere/inky` image, which contains duplicate 
> layer ids.
> 
> 
> Thanks,
> 
> Gilbert Song
> 
>

Reply via email to