Re: Review Request 45513: Fixed mount point check in port mapping isolator.

[Jie Yu]

> On March 31, 2016, 12:14 a.m., Jie Yu wrote:
> > src/slave/containerizer/mesos/isolators/network/port_mapping.cpp, lines 
> > 1901-1909
> > 
> >
> > Thanks for the patch!
> > 
> > Actually, I am working on a patch already. There are more issues here. 
> > We need to make sure /var/run/netns (or /run/netns) is a shared mount in 
> > its own mount peer group so that we don't create extra reference to the 
> > mount when launching a container (which will clone a new mount namespace, 
> > thus adding a reference to the mount). That'll make sure the rmdir of the 
> > mount points will succeeds after umount (othewise, we might get EBUSY).

My patch posted here:
https://reviews.apache.org/r/45520

Can you take a look (and test it in your environment)?


- Jie


---
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/45513/#review126229
---


On March 31, 2016, 12:04 a.m., Cong Wang wrote:
> 
> ---
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/45513/
> ---
> 
> (Updated March 31, 2016, 12:04 a.m.)
> 
> 
> Review request for mesos, Ian Downes and Jie Yu.
> 
> 
> Repository: mesos
> 
> 
> Description
> ---
> 
> On Fedora, /var/run is a symlink to /run, which causes the mount point check 
> code fail to detect it therefore keeps mounting it each time when we restart 
> the slave.
> 
> 
> Diffs
> -
> 
>   src/slave/containerizer/mesos/isolators/network/port_mapping.cpp 
> 323c84a3d960a196d8ba87f753814e9d43a07957 
> 
> Diff: https://reviews.apache.org/r/45513/diff/
> 
> 
> Testing
> ---
> 
> Manually start mesos slave and verify there is no more leak in /proc/mounts.
> 
> 
> Thanks,
> 
> Cong Wang
> 
>

Re: Review Request 45513: Fixed mount point check in port mapping isolator.

[Jie Yu]

---
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/45513/#review126229
---




src/slave/containerizer/mesos/isolators/network/port_mapping.cpp (lines 1901 - 
1909)


Thanks for the patch!

Actually, I am working on a patch already. There are more issues here. We 
need to make sure /var/run/netns (or /run/netns) is a shared mount in its own 
mount peer group so that we don't create extra reference to the mount when 
launching a container (which will clone a new mount namespace, thus adding a 
reference to the mount). That'll make sure the rmdir of the mount points will 
succeeds after umount (othewise, we might get EBUSY).


- Jie Yu


On March 31, 2016, 12:04 a.m., Cong Wang wrote:
> 
> ---
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/45513/
> ---
> 
> (Updated March 31, 2016, 12:04 a.m.)
> 
> 
> Review request for mesos, Ian Downes and Jie Yu.
> 
> 
> Repository: mesos
> 
> 
> Description
> ---
> 
> On Fedora, /var/run is a symlink to /run, which causes the mount point check 
> code fail to detect it therefore keeps mounting it each time when we restart 
> the slave.
> 
> 
> Diffs
> -
> 
>   src/slave/containerizer/mesos/isolators/network/port_mapping.cpp 
> 323c84a3d960a196d8ba87f753814e9d43a07957 
> 
> Diff: https://reviews.apache.org/r/45513/diff/
> 
> 
> Testing
> ---
> 
> Manually start mesos slave and verify there is no more leak in /proc/mounts.
> 
> 
> Thanks,
> 
> Cong Wang
> 
>