----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/32219/#review77335 -----------------------------------------------------------
Ship it! src/slave/containerizer/isolators/network/port_mapping.cpp <https://reviews.apache.org/r/32219/#comment125246> This looks good to me. As we discussed, please add a comment about the fact that the rate set above is the max deque rate from the qdisc. Packets will be buffered at the qdisc and maybe dropped if the qdisc buffer is full. We choose to use fq_codel because it has a better flow management and a larger default buffer than pfifo_fast. - Jie Yu On March 18, 2015, 9:38 p.m., Cong Wang wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/32219/ > ----------------------------------------------------------- > > (Updated March 18, 2015, 9:38 p.m.) > > > Review request for mesos, Chi Zhang, Ian Downes, and Jie Yu. > > > Bugs: MESOS-2514 > https://issues.apache.org/jira/browse/MESOS-2514 > > > Repository: mesos > > > Description > ------- > > When we enable bandwidth cap, htb is used on egress side inside containers, > however, the default leaf qdisc for a htb class is still pfifo_fast, which is > known to have buffer bloat. Change the default leaf qdisc to fq_codel too. > > > Diffs > ----- > > src/slave/containerizer/isolators/network/port_mapping.cpp > e5e6fa66e4032ffc6b128b5b6adf1d2e551b8f1f > > Diff: https://reviews.apache.org/r/32219/diff/ > > > Testing > ------- > > Manually started a container and fq_codel qdisc was successfully created. > > > Thanks, > > Cong Wang > >
