OpenShift has some cAdvisor packages in its import chain. The init() here
invokes Dbus start even when cAdvisor is not actively running or being
used (in our packages). This change makes the systemd check and dbus
initialization and connection lazy, occuring only the first time someone
invokes UseSystemd().
This allows us to handle the case where there are no factories as
happens when testing in a non-container environment. We will still serve
the machine information that is available.
/storage returns {device, mountpoint, capacity, usage} for all filesystems.
In addition, it also detect and applies label for each filesystem - currently two - "root", "docker-images".
/storage/<label> returns info about the filesystem with specific label. eg. /storage/root returns info for root filesystem.
This should make us more robust in the face of failure (at the cost of
making the failures less prominent). We allow GetStats() to return an
error and a partial result. We will process the result and report the
error.
Fixes#306.
The stats are only populated when cAdvisor is running outside network namespaces.
We'll add a different backend to retrieve the same data from within namespaces.
This adds an Exists() interface to detect when the container is dead.
Before reporting an update error we check is Exists() is true.
Some documentation was added as well.
We used to only register them on the first hierarchy that was created (I
think this was unintentional). This caused some weird edgecases where
we'd try to delete a watch event we didn't create. It is an error we
ignore today (since we fix it in < 60s) but delays our destruction of
the container.
We used to include it with the trailing "\n" which gave us parsing
errors in the UI (and the reason why they didn't show up in any of those
containers).
This is done by introducting the concept of "namespaces" of container
names. The aliases of a container are under this namespace. Namespace
names are of the form:
//<namespace>/<alias>
This allows us to (within cAdvisor) query all docker containers as
//docker regardless of whether this is a systemd or a non-systemd system.
This does break our ability to handle Docker aliases with the /container
endpoint. I think this is acceptable as our support there was not
consistent between system types.
