Issue 287: Deleting SegmentStore Services after Scaledown

[SrishT]

Signed-off-by: SrishT [email protected]

Change log description

When we deploy PravegaCluster with external access enabled and scale down the segmentstore to a lower instance count, the service entries for the segmentstore were not getting deleted.

Purpose of the change

Fixes #287

How to verify it

To reproduce the issue, deployed a sample Pravega Cluster with 3 segmentstore replicas and with external access enabled. Updated the number of segmentstore replicas in the spec to 1 and observed that while the extra segmentstore replicas were getting deleted, the extra segmentstore services were left behind.

$ kubectl get all
NAME                                                       READY   STATUS    RESTARTS   AGE
pod/pravega-bookie-0                                       1/1     Running   0          164m
pod/pravega-bookie-1                                       1/1     Running   0          164m
pod/pravega-bookie-2                                       1/1     Running   0          164m
pod/pravega-pravega-controller-fc6857f5-9429m              1/1     Running   0          164m
pod/pravega-pravega-segmentstore-0                         1/1     Running   0          164m

NAME                                            TYPE           CLUSTER-IP       EXTERNAL-IP     PORT(S)                          AGE
service/pravega-bookie-headless                 ClusterIP      None             <none>          3181/TCP                         164m
service/pravega-pravega-controller              LoadBalancer   10.100.200.216   10.247.113.80   10080:32657/TCP,9090:30330/TCP   164m
service/pravega-pravega-segmentstore-0          LoadBalancer   10.100.200.185   10.247.113.86   12345:30313/TCP                  164m
service/pravega-pravega-segmentstore-1          LoadBalancer   10.100.200.194   10.247.113.86   12345:30313/TCP                  164m
service/pravega-pravega-segmentstore-2          LoadBalancer   10.100.200.196   10.247.113.86   12345:30313/TCP                  164m
service/pravega-pravega-segmentstore-headless   ClusterIP      None             <none>          12345/TCP                        164m

Created a new operator image with the fix and tested the same scenario. This time, the extra segmentstore service entries were also deleted, as expected.

$ kubectl get all
NAME                                                       READY   STATUS    RESTARTS   AGE
pod/pravega-bookie-0                                       1/1     Running   0          11m
pod/pravega-bookie-1                                       1/1     Running   0          11m
pod/pravega-bookie-2                                       1/1     Running   0          11m
pod/pravega-pravega-controller-fc6857f5-4dnwg              1/1     Running   0          11m
pod/pravega-pravega-segmentstore-0                         1/1     Running   0          11m

NAME                                            TYPE           CLUSTER-IP       EXTERNAL-IP     PORT(S)                          AGE
service/pravega-bookie-headless                 ClusterIP      None             <none>          3181/TCP                         11m
service/pravega-pravega-controller              LoadBalancer   10.100.200.142   10.247.113.80   10080:30692/TCP,9090:30030/TCP   11m
service/pravega-pravega-segmentstore-0          LoadBalancer   10.100.200.118   10.247.113.86   12345:30313/TCP                  11m
service/pravega-pravega-segmentstore-headless   ClusterIP      None             <none>          12345/TCP                        11m

[pbelgundi]

Travis build is failing. please fix the same.
Please add unit tests to cover the new code.
Some other minor changes suggested as well.

[pbelgundi]

LGTM