some pods have no containers about cluster-insight HOT 3 CLOSED

The list of containers obtained from Docker on this node using:
% curl http://localhost:4243/containers/json

is:
[
...
,{"Command":""/bin/sh -c '/usr/sbin/td-agent "$FLUENTD_ARGS" \u003e /var/log/td-agent/td-agent.log'"","Created":1428433466,"Id":"00be5cd66a8f9dc82c9ec00071b6fbc17232e8c977db252eccfc4f2aeae21be3","Image":"kubernetes/fluentd-elasticsearch:1.3","Names":["/k8s_fluentd-es.2a803504_fluentd-to-elasticsearch-kubernetes-minion-a7lt.c.gce-monitoring.internal_default_c5973403e9c9de201f684c38aa8a7588_4dfe38b6"],"Ports":[],"Status":"Up 44 hours"}
,{"Command":"/pause","Created":1428433465,"Id":"7965cae79197413216add4d418f3b5a218ae3b240101fc69d57c2898abcf98dc","Image":"kubernetes/pause:go","Names":["/k8s_POD.7c16d80d_fluentd-to-elasticsearch-kubernetes-minion-a7lt.c.gce-monitoring.internal_default_c5973403e9c9de201f684c38aa8a7588_417b0b4c"],"Ports":[],"Status":"Up 44 hours"}

This container appears in the output of the /cluster/resources/containers API.

Interestingly, the value of the "Hostname" attribute in the container information is truncated to the first 64 characters:
"Hostname": "fluentd-to-elasticsearch-kubernetes-minion-a7lt.c.gce-monitorin",
This truncation is probably a red herring.

from cluster-insight.

This truncation is the source of the problem. The "Hostname" attribute actually contains the pod name and not a true host name. We try to match this in the container with the pod name. Since it is truncated, we do not match the the pod name, and this pod remains unreported.

from cluster-insight.

Fixed by pull request #39.

from cluster-insight.