[feedback] Chronos/Netdata

[kopax]

Hi panteras users,

@cookandy , I have a question regarding netdata, how do you found it for monitoring and logging ? The leading docker monitoring opensource service has changed.

• https://stackshare.io/stackups/grafana-vs-netdata
• https://grafana.com/
• https://blog.eleven-labs.com/fr/monitorer-ses-containers-docker/

In a much more advanced state than netdata, should we consider renewing these tools?

Also, I have found Chronos as a complicated/unfinished/heavy scheduling tool, we by default decided to stop using it because of it's reliability.

Are you people using it? I am using classic crontab for scheduling but this require SSHing to the server for editing the crontab.

[cookandy]

I must admit I haven't used netdata in quite some time. Most of my production servers are monitored by Scout and other monitoring tools. The container metrics are either written to statsd or grabbed by prometheus.

With regards to Chronos, I've also found it quite buggy and usually just use cron as well.

[sielaq]

We use Graphite or Prometheus and for viewing Grafana. Netdata is by default disabled.

Regarding Chronos - we use it occasionally (one usecase)

[kopax]

Hi and Happy new years! As @sielaq use Grafana and @cookandy doesn't use it anymore:

• should I consider removing netdata and install grafana instead in the Dockerfile?

Also, I'd like to ask you both about data persistence accross cluster. So far An app with data must be started with only one instance and using fixed hostname when starting the container in marathon.

• Is there any driver I should use with marathon ?
• I used to test glusterfs and some over but all where unsatisfying, I was finally considering to build my own driver on top of nfs. Did the stack change recently ?

[sielaq]

should I consider removing netdata and install grafana instead in the Dockerfile?

you don't have to start netdata. its the same with 100s of other apps you have in the image but you don use them. Moreover installing grafana will not help you much - you need a source of data (graphite/prometheus...) for that you have different project - don't try to put all the apps into one container -not worth to.

Is there any driver I should use with marathon ?

How to combine Docker/Mesos/Marathon with Grafana+graphite: you can use collectd: https://github.com/bogus-py/docker-collectd-plugin (for docker vs marathon match) https://github.com/rayrod2030/collectd-mesos (for mesos statistics ) etc.. and use build in write_graphite plugin

If you want to get metrics from application (service related) you can do in 2 ways: push or pull:

• with push - app gonna push directly to graphite
• with pull - app should expose endpoint with metrics /metrics and for ex. prometheus should read and collect that

If you choose prometheus you can use lots of exportres https://prometheus.io/docs/instrumenting/exporters/

so, no adding more stuff to PanteraS is needed - just use additional projects and integrate them as you wish.

[kopax]

so, no adding more stuff to PanteraS is needed - just use additional projects and integrate them as you wish.

Ok. I will close this then.

I've just read the overview of prometheus and it look promising. I think I've read gitlab support metrics integration from prometheus. How do you use the exporteds metrics ? Is it grafana job ?

[sielaq]

grafana is only for drawing / plotting / alerting based on multiple sources of data the data source can be: Graphite, Prometheus, Influx etc... you can use collectd (for graphite) or prometheus exporters (for prometheus)