This causes a side-effect for us where our logging program which is file path based, has a hard time matching up stderr / stdout and other logs to the applications that are generating them. I've been meaning to write a https://mesos.apache.org/documentation/latest/logging/ ContainerLogger that can make sure the right metadata is populated, but haven't found the eng time to do so.
On Fri, Feb 24, 2017 at 1:15 PM Jie Yu <yujie....@gmail.com> wrote: > The problem referred in the ticket does not exist any more. We now use > Mesos generated container ID as the cgroup path (since 0.18?). > > That reminds me that the nested container ID might have this restriction > because it's generated by the executor. We should add a validation in the > agent. > > - Jie > > On Thu, Feb 23, 2017 at 6:14 PM, Cody G <codyhg...@gmail.com> wrote: > > > It came to my attention via this owner-less ticket > > <https://issues.apache.org/jira/browse/MESOS-691> that Mesos has issues > > with task ids that are more than 255 characters. Is this still a problem? > > > > My reason for asking is that Aurora places restrictions on role / job > name > > based on the assumption that Mesos will crash if the task id is too long. > > > > Thanks, > > Cody > > >
