Naming

[evilezh]

Make proper naming we have /id1/id2/id3/id4/<dynamic id>/<sub id>/<metric> /id1/id2/id3/id4/id5/<sub id>/<metric> better would be

/id1/id2/id3/id4/id5/<dynamic id>/<sub id>/<metric> /id1/id2/id3/id4/id5/<sub id>/<metric>

So - idea is to have for all namespaces initial id's same amount - so it is easier to classify and identify.

if we have as it is .. then I need to guess .. is that framework id which will come or something else. Currently you have options ..

1. identify well known id5 .. and if it does not match .. assume it is framework id
2. try some regex match .. Both cases might file, when you add new stats, which result of failing parser code.

I'm sending stats to kafka+json .. so it is quite funny to decode now. :)

[evilezh]

Here is exact example: /intel/mesos/master/<executor>/offered_resources/name /intel/mesos/master/<executor>/resources/name /intel/mesos/master/<executor>/user_resources/name

/intel/mesos/master/registrar/state_store_ms/ /intel/mesos/master/registrar/ /intel/mesos/master/system /intel/mesos/master/master

if we could have something like: /intel/mesos/master/executors/<executor>/offered_resources/name

Something similar with slave/agent...

[ghost]

There are two things at play here:

1. Per-executor metrics
2. Mesos system metrics

In the case of executor metrics, yes we can add executors to the namespace with very little effort.

In the case of Mesos system metric namespace components (for example: /intel/mesos/master/cpus_total vs. /intel/mesos/master/registrar/state_store_ms/p90), we actually don't have any control over the registrar/state_store_ms/p90 part; the forward slashes (/) are actually being introduced by Mesos[1], and we build the Snap metric namespace with them.

[1] http://mesos.apache.org/documentation/latest/monitoring/