fix: enabling deleted grafana datasources to be recreated #96
Conversation
prashbnair
force-pushed
the
grafana-ds
branch
from
eb2bd7a
to
cd13e13
Compare
|
@sthaha do you think we want an e2e test here? |
prashbnair
force-pushed
the
grafana-ds
branch
from
cd13e13
to
8a02154
Compare
I agree, we should test if operator restores datasource if it is modified or deleted. I suspect that given then current code that it might crash if the datasource is deleted. |
Yes this has been tested. |
Can we do this in a separate PR? |
Sure thing :-) |
|
@prashbnair , Could you please also ensure that deletion a namespace that contains monitoring-stacks also cleans up their associated grafana-datasources ? |
|
As discussed, it would be nice to have all reconciliation issues sorted in a single or multiple PRs |
prashbnair
force-pushed
the
grafana-ds
branch
from
8a02154
to
3b587e0
Compare
|
@sthaha I will address the clean up of grafana data sources when the monitoring stack or namespace is deleted as a separate PR. |
prashbnair
force-pushed
the
grafana-ds
branch
from
3b587e0
to
4087a83
Compare
|
|
||
| if err != nil { | ||
| log.V(3).Info("grafana data source CRD is not defined") | ||
| return err |
There was a problem hiding this comment.
I think we should follow same the implementation in @fpetkovski 's PR ... https://github.com/rhobs/monitoring-stack-operator/pull/103/files#diff-606d75b2cccce3c13550d55e0c1b27bb1c1fad0431b4f16d466e41dd4e0c801dR355
I.E. function should return requeue (bool) and error
func setWatch() (requeue bool, err error){
if err = listDataSources(); err != nil
return true, nil
}
err = setWatch()
return false, err
}
So that the caller (Reconcile func) can ...
if requeue, err := r.setGrafanaDatasourceWatch(); requeue || err != nil {
if err {
return ctrl.Result{}, err
}
return ctrl.Result{RequeueAfter: 10 * time.Second}, nil
}There was a problem hiding this comment.
All the patchers in this reconcile function return an error , I have followed the same model.
There was a problem hiding this comment.
This part is different from the rest of the patchers. This isn't a patcher to start with :-)
More importantly ...
We need the ability to requeue without error if GrafanaDataSource CRD is not created and to log real errors when setting up a watch fails.
|
|
||
| f.K8sClient.Delete(context.Background(), &grafanaDS) | ||
|
|
||
| f.GetResourceWithRetry(t, datasourceName, "monitoring-stack-operator", &grafanaDS) |
There was a problem hiding this comment.
I think we should make use of AssertResourceEventuallyExists here as well
prashbnair
force-pushed
the
grafana-ds
branch
from
4087a83
to
f6dde2d
Compare
prashbnair
force-pushed
the
grafana-ds
branch
2 times, most recently
from
2e1e30c
to
e094bcb
Compare
| assert.NilError(t, err, "failed to create a monitoring stack") | ||
|
|
||
| grafanaDS := grafanav1alpha1.GrafanaDataSource{} | ||
|
|
There was a problem hiding this comment.
nit: do we need a blank line?
There was a problem hiding this comment.
Good stuff @prashbnair
You may want to consider incorporating the nits
prashbnair
force-pushed
the
grafana-ds
branch
from
e094bcb
to
3027ffd
Compare
| if err != nil { | ||
| return ctrl.Result{}, err | ||
| } | ||
|
|
| @@ -82,7 +82,7 @@ func (f *Framework) AssertResourceEventuallyExists(name, namespace string, resou | |||
| } | |||
| return false, nil | |||
| }); err == wait.ErrWaitTimeout { | |||
| t.Fatal(fmt.Errorf("statefulset %s/%s was never created", namespace, name)) | |||
| t.Fatal(fmt.Errorf("resource %s/%s was never created", namespace, name)) | |||
prashbnair
force-pushed
the
grafana-ds
branch
from
3027ffd
to
857a8a3
Compare
Prior to this fix, grafana datasources which were deleted were not being recreated.
prashbnair
force-pushed
the
grafana-ds
branch
from
857a8a3
to
b6506ff
Compare
No description provided.