public_git/postgres-operator
1
0
Fork 0
mirror of https://github.com/zalando/postgres-operator.git synced 2026-09-30 16:27:28 +02:00
Code Issues Projects Releases Wiki Activity

Implement per-cluster maintenance window for Postgres automatic upgrade (#2710)

Browse Source 
* implement maintenance window for major version upgrade 
* e2e test: fix major version upgrade test and extend with the time window
* unit test: add iteration to test isInMaintenanceWindow
* UI: show the window and enable edit via UI
This commit is contained in:
Ida Novindasari
2024-08-09 14:07:35 +02:00
committed by GitHub
parent ce15d10aa3
commit e6ae9e3772
10 changed files with 219 additions and 24 deletions
Download Patch File Download Diff File Expand all files Collapse all files
e2e/tests/k8s_api.py
-2
View File
@@ -218,7 +218,6 @@ class K8s:
pod_phase = 'Failing over'
new_pod_node = ''
pods_with_update_flag = self.count_pods_with_rolling_update_flag(labels, namespace)
while (pod_phase != 'Running') or (new_pod_node not in failover_targets):
pods = self.api.core_v1.list_namespaced_pod(namespace, label_selector=labels).items
if pods:
@@ -525,7 +524,6 @@ class K8sBase:
pod_phase = 'Failing over'
new_pod_node = ''
pods_with_update_flag = self.count_pods_with_rolling_update_flag(labels, namespace)
while (pod_phase != 'Running') or (new_pod_node not in failover_targets):
pods = self.api.core_v1.list_namespaced_pod(namespace, label_selector=labels).items
if pods:
e2e/tests/test_e2e.py
+79 -14
View File
@@ -14,6 +14,7 @@ from kubernetes.client.rest import ApiException
SPILO_CURRENT = "registry.opensource.zalan.do/acid/spilo-16-e2e:0.1"
SPILO_LAZY = "registry.opensource.zalan.do/acid/spilo-16-e2e:0.2"
SPILO_FULL_IMAGE = "ghcr.io/zalando/spilo-16:3.2-p3"
def to_selector(labels):
@@ -115,6 +116,7 @@ class EndToEndTestCase(unittest.TestCase):
configmap = yaml.safe_load(f)
configmap["data"]["workers"] = "1"
configmap["data"]["docker_image"] = SPILO_CURRENT
configmap["data"]["major_version_upgrade_mode"] = "full"
with open("manifests/configmap.yaml", 'w') as f:
yaml.dump(configmap, f, Dumper=yaml.Dumper)
@@ -1181,31 +1183,94 @@ class EndToEndTestCase(unittest.TestCase):
self.eventuallyEqual(lambda: len(k8s.get_patroni_running_members("acid-minimal-cluster-0")), 2, "Postgres status did not enter running")
@timeout_decorator.timeout(TEST_TIMEOUT_SEC)
@unittest.skip("Skipping this test until fixed")
def test_major_version_upgrade(self):
k8s = self.k8s
result = k8s.create_with_kubectl("manifests/minimal-postgres-manifest-12.yaml")
self.eventuallyEqual(lambda: k8s.count_running_pods(labels="application=spilo,cluster-name=acid-upgrade-test"), 2, "No 2 pods running")
self.eventuallyEqual(lambda: k8s.get_operator_state(), {"0": "idle"}, "Operator does not get in sync")
"""
Test major version upgrade
"""
def check_version():
p = k8s.patroni_rest("acid-upgrade-test-0", "")
version = p.get("server_version", 0) // 10000
return version
pg_patch_version = {
k8s = self.k8s
cluster_label = 'application=spilo,cluster-name=acid-upgrade-test'
with open("manifests/minimal-postgres-manifest-12.yaml", 'r+') as f:
upgrade_manifest = yaml.safe_load(f)
upgrade_manifest["spec"]["dockerImage"] = SPILO_FULL_IMAGE
with open("manifests/minimal-postgres-manifest-12.yaml", 'w') as f:
yaml.dump(upgrade_manifest, f, Dumper=yaml.Dumper)
k8s.create_with_kubectl("manifests/minimal-postgres-manifest-12.yaml")
self.eventuallyEqual(lambda: k8s.count_running_pods(labels=cluster_label), 2, "No 2 pods running")
self.eventuallyEqual(lambda: k8s.get_operator_state(), {"0": "idle"}, "Operator does not get in sync")
self.eventuallyEqual(check_version, 12, "Version is not correct")
master_nodes, _ = k8s.get_cluster_nodes(cluster_labels=cluster_label)
# should upgrade immediately
pg_patch_version_14 = {
"spec": {
"postgres": {
"postgresql": {
"version": "14"
}
}
}
k8s.api.custom_objects_api.patch_namespaced_custom_object(
"acid.zalan.do", "v1", "default", "postgresqls", "acid-upgrade-test", pg_patch_version)
"acid.zalan.do", "v1", "default", "postgresqls", "acid-upgrade-test", pg_patch_version_14)
self.eventuallyEqual(lambda: k8s.get_operator_state(), {"0": "idle"}, "Operator does not get in sync")
def check_version_14():
p = k8s.get_patroni_state("acid-upgrade-test-0")
version = p["server_version"][0:2]
return version
# should have finish failover
k8s.wait_for_pod_failover(master_nodes, 'spilo-role=replica,' + cluster_label)
k8s.wait_for_pod_start('spilo-role=master,' + cluster_label)
k8s.wait_for_pod_start('spilo-role=replica,' + cluster_label)
self.eventuallyEqual(check_version, 14, "Version should be upgraded from 12 to 14")
self.eventuallyEqual(check_version_14, "14", "Version was not upgrade to 14")
# should not upgrade because current time is not in maintenanceWindow
current_time = datetime.now()
maintenance_window_future = f"{(current_time+timedelta(minutes=60)).strftime('%H:%M')}-{(current_time+timedelta(minutes=120)).strftime('%H:%M')}"
pg_patch_version_15 = {
"spec": {
"postgresql": {
"version": "15"
},
"maintenanceWindows": [
maintenance_window_future
]
}
}
k8s.api.custom_objects_api.patch_namespaced_custom_object(
"acid.zalan.do", "v1", "default", "postgresqls", "acid-upgrade-test", pg_patch_version_15)
self.eventuallyEqual(lambda: k8s.get_operator_state(), {"0": "idle"}, "Operator does not get in sync")
# should have finish failover
k8s.wait_for_pod_failover(master_nodes, 'spilo-role=master,' + cluster_label)
k8s.wait_for_pod_start('spilo-role=master,' + cluster_label)
k8s.wait_for_pod_start('spilo-role=replica,' + cluster_label)
self.eventuallyEqual(check_version, 14, "Version should not be upgraded")
# change the version again to trigger operator sync
maintenance_window_current = f"{(current_time-timedelta(minutes=30)).strftime('%H:%M')}-{(current_time+timedelta(minutes=30)).strftime('%H:%M')}"
pg_patch_version_16 = {
"spec": {
"postgresql": {
"version": "16"
},
"maintenanceWindows": [
maintenance_window_current
]
}
}
k8s.api.custom_objects_api.patch_namespaced_custom_object(
"acid.zalan.do", "v1", "default", "postgresqls", "acid-upgrade-test", pg_patch_version_16)
self.eventuallyEqual(lambda: k8s.get_operator_state(), {"0": "idle"}, "Operator does not get in sync")
# should have finish failover
k8s.wait_for_pod_failover(master_nodes, 'spilo-role=replica,' + cluster_label)
k8s.wait_for_pod_start('spilo-role=master,' + cluster_label)
k8s.wait_for_pod_start('spilo-role=replica,' + cluster_label)
self.eventuallyEqual(check_version, 16, "Version should be upgraded from 14 to 16")
@timeout_decorator.timeout(TEST_TIMEOUT_SEC)
def test_persistent_volume_claim_retention_policy(self):