Uploaded image for project: 'DC/OS'
  1. DC/OS
  2. DCOS_OSS-4320

dcos-dns incorrectly deletes records in PODs when one container exits.

    Details

      Description

      The following POD definition can be used to reproduce this bug in DC/OS 1.12:

      {
        "id": "/ucr/dns-bug",
        "version": "2018-10-13T23:53:29.644Z",
        "containers": [
          {
            "name": "container-1",
            "resources": {
              "cpus": 0.1,
              "mem": 128,
              "disk": 0,
              "gpus": 0
            },
            "exec": {
              "command": {
                "shell": "sleep 1000000"
              }
            },
            "image": {
              "kind": "DOCKER",
              "id": "alpine"
            }
          },
          {
            "name": "container-2",
            "resources": {
              "cpus": 0.1,
              "mem": 128,
              "disk": 0,
              "gpus": 0
            },
            "exec": {
              "command": {
                "shell": "sleep 60"
              }
            },
            "image": {
              "kind": "DOCKER",
              "id": "alpine"
            }
          }
        ],
        "networks": [
          {
            "name": "dcos",
            "mode": "container"
          }
        ],
        "scaling": {
          "instances": 1,
          "kind": "fixed"
        },
        "scheduling": {
          "placement": {
            "constraints": []
          }
        },
        "executorResources": {
          "cpus": 0.1,
          "mem": 32,
          "disk": 10
        },
        "volumes": [],
        "fetch": []
      }
      

      Initially, it generates the correct DNS entries:

      {"host":"9.0.1.2","name":"dns-bug-ucr.marathon.autoip.dcos.thisdcos.directory","rtype":"A"},
      {"host":"9.0.1.2","name":"dns-bug-ucr.marathon.containerip.dcos.thisdcos.directory","rtype":"A"},
      

      Once the second container exits, after 60seconds in my example, the DNS records get deleted.

      I was not able to reproduce this behaviour on DC/OS 1.11 - hence this is a regression.

        Attachments

          Activity

            People

            • Assignee:
              sergeyurbanovich Sergey Urbanovich
              Reporter:
              till Till Toenshoff
              Team:
              Networking Team
              Watchers:
              Deepak Goel, Jan-Philip Gehrcke, Mergebot, Orlando Hohmeier, Sergey Urbanovich, Till Toenshoff
            • Watchers:
              6 Start watching this issue

              Dates

              • Created:
                Updated:
                Resolved: