> On Oct. 27, 2015, 5:51 p.m., Jie Yu wrote:
> > src/slave/containerizer/isolators/network/port_mapping.cpp, lines 1403-1404
> > <https://reviews.apache.org/r/39415/diff/1/?file=1100525#file1100525line1403>
> >
> >     Can you explain more? What if the slave restarts after creating the 
> > qdisc, the slave will flap afterwards?

Hmm, if there any way to tell if this is a recover or normal start?

The problem I want to solve is:

1) create an HTB qdisc on root;
2) start mesos slave with this flag, which will try to create fq_codel on root 
too;
3) No error, because we ignore if root qdisc exists before us
4) But packets get dropped because our filters created for root qdisc are not 
supposed to be used for HTB

The silent drop is definitely not acceptable...


- Cong


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/39415/#review104187
-----------------------------------------------------------


On Oct. 18, 2015, 12:22 a.m., Cong Wang wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/39415/
> -----------------------------------------------------------
> 
> (Updated Oct. 18, 2015, 12:22 a.m.)
> 
> 
> Review request for mesos, Ian Downes and Jie Yu.
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> When we enable --egress_unique_flow_per_container, we need to install an 
> fq_codel qdisc on root, but if a qdisc already exists on root, we should not 
> continue since it could be not fq_codel at all, so just exit with error.
> 
> 
> Diffs
> -----
> 
>   src/slave/containerizer/isolators/network/port_mapping.cpp 
> e6bb75e6f5ba48a0c4cf6dd8f353e5f5843d0eef 
> 
> Diff: https://reviews.apache.org/r/39415/diff/
> 
> 
> Testing
> -------
> 
> Manual tests, with a pre-installed HTB qdisc and without.
> 
> 
> Thanks,
> 
> Cong Wang
> 
>

Reply via email to