Open Stack

On 07/01/2021 13:20, Golasowski Martin wrote:
> Dear All,
> I would like to know which monitoring solution is currently supported by 
> OSA? We are operating a small cloud (~ 6 nodes) and we are interested in 
> collecting performance metrics, events and logs.
> 
> So, as far as I know, the official OSA solution is ceilometer/aodh/panko 
> with Gnocchi as DB backend. However Gnocchi project seems abandoned at 
> the moment and the grafana plugin is not compatible with latest Grafana.
> 
> Then there is solution based on collectd with this plugin 
> (https://github.com/signalfx/collectd-openstack 
> <https://github.com/signalfx/collectd-openstack>) with Graphite or 
> InfluxDB as backend. This supports only performance metrics and not the 
> events.
> 
> Then there are also some Prometheus exporters available, again, metrics 
> only.
> 
> What do you guys use these days? What would you recommend?
> 

Hi there,

with my telemetry hat on: we're working on the gnocchi issue, but 
gnocchi is only a metrics store anyways. Personally, I wouldn't want to 
store any events in panko. If you use such things like autoscaling for 
instances, you definitely want gnocchi and aodh.

With my collectd hat on: collectd supports collecting and sending 
metrics and events to multiple write endpoints. It is not designed to 
collect additional metadata, such as project or user data. You'll mostly 
get infrastructure related data (from baremetal nodes).
The con side with using graphite or influxdb in the Open Source variant 
is, that you don't get HA.

There is the Service Telemetry Framework[1], but it is integrated with 
TripleO, not with OSA, it uses both collectd and ceilometer for 
collection; metrics are stored in prometheus, events in elasticsearch, 
which is also used for log aggregation. I am unsure if this solution is 
not a bit too heavy for your use case.


The best interest (in this community here): put some man-power on gnocchi.

Matthias


[1] https://infrawatch.github.io/documentation/