> On Sept. 6, 2016, 6:54 p.m., Avinash sridharan wrote: > > src/health-check/health_checker.cpp, line 520 > > <https://reviews.apache.org/r/51607/diff/1/?file=1490612#file1490612line520> > > > > This more of a question: > > Why is the --host hard coded to the `DEFAULT_DOMAIN`. What if the > > container is binding to a specific IP address, instead of any ADDR_ANY? > > Also, in case of the docker containerizer, the `tcpConnect` binary is going > > to be running in the host network namespace so connecting to localhost > > might work for cases where the port-mapping is setup for the containers, > > but don't think it will work for cases where the container (docker > > container) has no port mapping but has it own network namespace. This will > > happen say with Calico or even the DC/OS overlay. > > Alexander Rukletsov wrote: > We decided that for simplicity. Is it always the case that the one can > tell what ip the task will have before launching the task? Probably no, and > we'll end up using discovery info for that task. Localhost sounds like a > minimal solution, but yes, the container should bind to all interfaces. > > In the docker case the `tcpConnect` binary will be running in the > container's network namespace, so it should be fine.
I do believe that there will be quite a few cases where containers will end up binding to a specific IP address. I think to make this more robust we should look at all interfaces in a given network namespace and do health check on all the IP addresses for a given endpoint. This shouldn't be that hard to do. Generally you will see only two IP addresses, the lo and the veth on the container. I think doing this will save us a lot of heartache in the future. - Avinash ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/51607/#review147887 ----------------------------------------------------------- On Oct. 4, 2016, 1:57 p.m., Alexander Rukletsov wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/51607/ > ----------------------------------------------------------- > > (Updated Oct. 4, 2016, 1:57 p.m.) > > > Review request for mesos, Avinash sridharan, Benjamin Mahler, Gastón Kleiman, > and haosdent huang. > > > Bugs: MESOS-6119 > https://issues.apache.org/jira/browse/MESOS-6119 > > > Repository: mesos > > > Description > ------- > > See summary. > > > Diffs > ----- > > src/docker/executor.cpp ab3f0473fdc9105d1c425f0dbe7b81c566d541e8 > src/health-check/health_checker.hpp > 392b4d5bd1e5831994b9366c1eb5a2911e19860f > src/health-check/health_checker.cpp > ea93132f2a5d4828c75005f102eddc4c3131599d > src/launcher/executor.cpp 8a1051b886a0c95d19ff370e5c77d9c4033c8b61 > > Diff: https://reviews.apache.org/r/51607/diff/ > > > Testing > ------- > > make check on Mac OS 10.11.6 > > > Thanks, > > Alexander Rukletsov > >
