[DPE-3039] Update Snap (#291)

* Update Snap

* Bump libs

* More verbose statuses for replicas

Author: dragomirp

lib/charms/data_platform_libs/v0/data_interfaces.py

@@ -320,7 +320,7 @@ def _on_topic_requested(self, event: TopicRequestedEvent):
 
 # Increment this PATCH version before using `charmcraft publish-lib` or reset
 # to 0 if you are raising the major API version
-LIBPATCH = 23
+LIBPATCH = 24
 
 PYDEPS = ["ops>=2.0.0"]
 
@@ -526,7 +526,16 @@ def get_content(self) -> Dict[str, str]:
         """Getting cached secret content."""
         if not self._secret_content:
             if self.meta:
-                self._secret_content = self.meta.get_content()
+                try:
+                    self._secret_content = self.meta.get_content(refresh=True)
+                except (ValueError, ModelError) as err:
+                    # https://bugs.launchpad.net/juju/+bug/2042596
+                    # Only triggered when 'refresh' is set
+                    msg = "ERROR either URI or label should be used for getting an owned secret but not both"
+                    if isinstance(err, ModelError) and msg not in str(err):
+                        raise
+                    # Due to: ValueError: Secret owner cannot use refresh=True
+                    self._secret_content = self.meta.get_content()
         return self._secret_content
 
     def set_content(self, content: Dict[str, str]) -> None:
@@ -1085,7 +1094,7 @@ def _delete_relation_secret(
         secret = self._get_relation_secret(relation.id, group)
 
         if not secret:
-            logging.error("Can't update secret for relation %s", str(relation.id))
+            logging.error("Can't delete secret for relation %s", str(relation.id))
             return False
 
         old_content = secret.get_content()
@@ -1827,7 +1836,8 @@ def _assign_relation_alias(self, relation_id: int) -> None:
 
         # We need to set relation alias also on the application level so,
         # it will be accessible in show-unit juju command, executed for a consumer application unit
-        self.update_relation_data(relation_id, {"alias": available_aliases[0]})
+        if self.local_unit.is_leader():
+            self.update_relation_data(relation_id, {"alias": available_aliases[0]})
 
     def _emit_aliased_event(self, event: RelationChangedEvent, event_name: str) -> None:
         """Emit an aliased event to a particular relation if it has an alias.
@@ -1914,6 +1924,9 @@ def _on_relation_created_event(self, event: RelationCreatedEvent) -> None:
 
         # Sets both database and extra user roles in the relation
         # if the roles are provided. Otherwise, sets only the database.
+        if not self.local_unit.is_leader():
+            return
+
         if self.extra_user_roles:
             self.update_relation_data(
                 event.relation.id,
@@ -2173,6 +2186,9 @@ def _on_relation_created_event(self, event: RelationCreatedEvent) -> None:
         """Event emitted when the Kafka relation is created."""
         super()._on_relation_created_event(event)
 
+        if not self.local_unit.is_leader():
+            return
+
         # Sets topic, extra user roles, and "consumer-group-prefix" in the relation
         relation_data = {
             f: getattr(self, f.replace("-", "_"), "")
@@ -2345,6 +2361,9 @@ def _on_relation_created_event(self, event: RelationCreatedEvent) -> None:
         """Event emitted when the OpenSearch relation is created."""
         super()._on_relation_created_event(event)
 
+        if not self.local_unit.is_leader():
+            return
+
         # Sets both index and extra user roles in the relation if the roles are provided.
         # Otherwise, sets only the index.
         data = {"index": self.index}

lib/charms/grafana_agent/v0/cos_agent.py

@@ -206,17 +206,15 @@ def __init__(self, *args):
 ```
 """
 
-import base64
 import json
 import logging
-import lzma
 from collections import namedtuple
 from itertools import chain
 from pathlib import Path
 from typing import TYPE_CHECKING, Any, Callable, ClassVar, Dict, List, Optional, Set, Union
 
 import pydantic
-from cosl import JujuTopology
+from cosl import GrafanaDashboard, JujuTopology
 from cosl.rules import AlertRules
 from ops.charm import RelationChangedEvent
 from ops.framework import EventBase, EventSource, Object, ObjectEvents
@@ -236,7 +234,7 @@ class _MetricsEndpointDict(TypedDict):
 
 LIBID = "dc15fa84cef84ce58155fb84f6c6213a"
 LIBAPI = 0
-LIBPATCH = 6
+LIBPATCH = 7
 
 PYDEPS = ["cosl", "pydantic < 2"]
 
@@ -251,31 +249,6 @@ class _MetricsEndpointDict(TypedDict):
 SnapEndpoint = namedtuple("SnapEndpoint", "owner, name")
 
 
-class GrafanaDashboard(str):
-    """Grafana Dashboard encoded json; lzma-compressed."""
-
-    # TODO Replace this with a custom type when pydantic v2 released (end of 2023 Q1?)
-    # https://github.com/pydantic/pydantic/issues/4887
-    @staticmethod
-    def _serialize(raw_json: Union[str, bytes]) -> "GrafanaDashboard":
-        if not isinstance(raw_json, bytes):
-            raw_json = raw_json.encode("utf-8")
-        encoded = base64.b64encode(lzma.compress(raw_json)).decode("utf-8")
-        return GrafanaDashboard(encoded)
-
-    def _deserialize(self) -> Dict:
-        try:
-            raw = lzma.decompress(base64.b64decode(self.encode("utf-8"))).decode()
-            return json.loads(raw)
-        except json.decoder.JSONDecodeError as e:
-            logger.error("Invalid Dashboard format: %s", e)
-            return {}
-
-    def __repr__(self):
-        """Return string representation of self."""
-        return "<GrafanaDashboard>"
-
-
 class CosAgentProviderUnitData(pydantic.BaseModel):
     """Unit databag model for `cos-agent` relation."""
 
@@ -748,6 +721,10 @@ def metrics_jobs(self) -> List[Dict]:
                         "job_name": job["job_name"],
                         "metrics_path": job["path"],
                         "static_configs": [{"targets": [f"localhost:{job['port']}"]}],
+                        # We include insecure_skip_verify because we are always scraping localhost.
+                        # Even if we have the certs for the scrape targets, we'd rather specify the scrape
+                        # jobs with localhost rather than the SAN DNS the cert was issued for.
+                        "tls_config": {"insecure_skip_verify": True},
                     }
 
                 scrape_jobs.append(job)

src/cluster.py

@@ -43,6 +43,8 @@
 
 PG_BASE_CONF_PATH = f"{POSTGRESQL_CONF_PATH}/postgresql.conf"
 
+RUNNING_STATES = ["running", "streaming"]
+
 
 class NotReadyError(Exception):
     """Raised when not all cluster members healthy or finished initial sync."""
@@ -297,7 +299,7 @@ def are_all_members_ready(self) -> bool:
         # because sometimes there may exist (for some period of time) only
         # replicas after a failed switchover.
         return all(
-            member["state"] == "running" for member in cluster_status.json()["members"]
+            member["state"] in RUNNING_STATES for member in cluster_status.json()["members"]
         ) and any(member["role"] == "leader" for member in cluster_status.json()["members"])
 
     def get_patroni_health(self) -> Dict[str, str]:
@@ -366,7 +368,7 @@ def member_started(self) -> bool:
         except RetryError:
             return False
 
-        return response["state"] == "running"
+        return response["state"] in RUNNING_STATES
 
     @property
     def member_inactive(self) -> bool:
@@ -381,7 +383,7 @@ def member_inactive(self) -> bool:
         except RetryError:
             return True
 
-        return response["state"] not in ["running", "starting", "restarting"]
+        return response["state"] not in [*RUNNING_STATES, "starting", "restarting"]
 
     @property
     def member_replication_lag(self) -> str:

src/constants.py

@@ -32,7 +32,7 @@
 # Snap constants.
 PGBACKREST_EXECUTABLE = "charmed-postgresql.pgbackrest"
 POSTGRESQL_SNAP_NAME = "charmed-postgresql"
-SNAP_PACKAGES = [(POSTGRESQL_SNAP_NAME, {"revision": "87"})]
+SNAP_PACKAGES = [(POSTGRESQL_SNAP_NAME, {"revision": "89"})]
 
 SNAP_COMMON_PATH = "/var/snap/charmed-postgresql/common"
 SNAP_CURRENT_PATH = "/var/snap/charmed-postgresql/current"