[Metricbeat] Migrate Ceph cluster_disk to use ReporterV2 interface #10990
Conversation
sayden
added
Metricbeat
sayden
changed the title
|
Error in Kibana/status module seems unrelated |
sayden
force-pushed
the
migration/mb/reporterv2/ceph/cluster_disk
branch
from
a8df0bb to
d13577f
Compare
|
https://github.com/elastic/beats/blob/master/metricbeat/module/ceph/cluster_disk/_meta/data.json does not need to be regenerated (in other words, when you regenerate it, nothing changes) |
| f := mbtest.NewEventFetcher(t, getConfig()) | ||
| err := mbtest.WriteEvent(f, t) | ||
| if err != nil { | ||
| t.Skip("Skipping data generation test") |
There was a problem hiding this comment.
What a nice timing 😄 #10990 (comment) and more info here #10993 (comment)
| f := mbtest.NewEventFetcher(t, getConfig()) | ||
| err := mbtest.WriteEvent(f, t) | ||
| if err != nil { | ||
| t.Skip("Skipping data generation test") |
There was a problem hiding this comment.
CI test aren't launched with -data option but this is checked in mbtest.WriteEventsReporterV2 in line 42. An error can still be raised in line 36 if ceph is not started by compose, which was producing flaky tests.
The problem is that Ceph start with Compose is buggy and it was omitted consistently in all Metricsets
There was a problem hiding this comment.
I see. As soon as #10648 is merged we should switch over Ceph to generate the data file based on an http server as the current tests already use a http server.
Let's skip it for now but open a follow up issue to track:
- Implementation of data generation based on [Metricbeat] Simplify testing http Metricbeat modules #10648
- And investigation in how we can tests ceph against a "real" server like we do for all the other modules.
sayden
force-pushed
the
migration/mb/reporterv2/ceph/cluster_disk
branch
from
d86447c to
2401d09
Compare
sayden
force-pushed
the
migration/mb/reporterv2/ceph/cluster_disk
branch
from
ede70fe to
11031ce
Compare
Refer to #10774 for more info