Details
-
Bug
-
Status: Resolved
-
Major
-
Resolution: Fixed
-
None
-
None
Description
As the operator of a Mesos cluster, I would like every container/executor to run in a single PID namespace, so that a task cannot see what else is running on the same host.
The existing namespaces/pid isolator seems to provide this feature. However, it seems like it is leaking files. I have exactly one task running currently, but there are still left overs from earlier invocations
vagrant@aurora:~/aurora$ ls -l /var/run/mesos/pidns/
total 0
-rw-r--r-- 1 root root 0 Aug 26 20:30 32b6e4c7-3d22-47ed-a350-9eb929daa241
-rw-r--r-- 1 root root 0 Aug 26 20:30 7b812f00-4614-4016-a76c-ff78a175a1b0
-rw-r--r-- 1 root root 0 Aug 26 20:24 d501829e-7cf8-40fb-a895-0ad3416da7dc
-rw-r--r-- 1 root root 0 Aug 26 20:24 d56ca91f-eb72-426c-8bbb-f3239358a4ef
-r--r--r-- 1 root root 0 Aug 26 20:35 fef9a109-de52-45f3-ae41-171de6495705
Attachments
Issue Links
- is related to
-
MESOS-5893 mesos-executor should adopt and reap orphan child processes
- Resolved
-
MESOS-6180 Several tests are flaky, with futures timing out early
- Resolved
- relates to
-
MESOS-1094 Introduce pid namespace abstraction to subprocess
- Open
-
MESOS-2725 undocumented how to enable pid namespace isolation and shared filesystem isolation
- Accepted