feat: Add keepalived to Anvil (#17)

---------

Co-authored-by: Stamatis Katsaounis <stamatis.katsaounis@canonical.com>

README.md

@@ -31,6 +31,9 @@ ubuntu@infra1:~$ maas-anvil cluster bootstrap \
     --accept-defaults
 ```
 
+Note: You will be asked for a `virtual_ip` during installation of the HAProxy charm, if `accept-defaults` is omitted.
+Pass an empty value to disable it, or any valid IP to enable; the Keepalived charm will be installed to enable connecting to HA MAAS using the VIP.
+
 ### Add new nodes to the MAAS cluster
 
 ```bash
@@ -74,14 +77,6 @@ ubuntu@infra1:~$ maas-anvil cluster list
 ubuntu@infra1:~$ juju run maas-region/0 create-admin username=admin password=pass email=admin@maas.io ssh-import=lp:maasadmin
 ```
 
-### (Optional) Add VIP functionality through Keepalived charm
-
-```bash
-ubuntu@infra1:~$ juju deploy containers/keepalived
-ubuntu@infra1:~$ juju relate haproxy:juju-info keepalived:juju-info
-ubuntu@infra1:~$ juju config keepalived virtual_ip=<ip_address>
-```
-
 ### Charm documentation
 
 - MAAS Region: <https://charmhub.io/maas-region>

anvil-python/anvil/commands/haproxy.py

@@ -13,38 +13,69 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from typing import List
+import ipaddress
+import logging
+from typing import Any, List
 
+from rich.console import Console
 from sunbeam.clusterd.client import Client
 from sunbeam.commands.terraform import TerraformInitStep
+from sunbeam.jobs import questions
+from sunbeam.jobs.common import BaseStep
 from sunbeam.jobs.juju import JujuHelper
-from sunbeam.jobs.manifest import BaseStep
 from sunbeam.jobs.steps import (
     AddMachineUnitsStep,
     DeployMachineApplicationStep,
     RemoveMachineUnitStep,
 )
 
 from anvil.jobs.manifest import Manifest
-from anvil.provider.local.deployment import LocalDeployment
+
+LOG = logging.getLogger(__name__)
 
 APPLICATION = "haproxy"
 CONFIG_KEY = "TerraformVarsHaproxyPlan"
+HAPROXY_CONFIG_KEY = "TerraformVarsHaproxy"
+KEEPALIVED_CONFIG_KEY = "TerraformVarsKeepalived"
 HAPROXY_APP_TIMEOUT = 180  # 3 minutes, managing the application should be fast
 HAPROXY_UNIT_TIMEOUT = (
     1200  # 15 minutes, adding / removing units can take a long time
 )
 
 
+def keepalived_questions() -> dict[str, questions.PromptQuestion]:
+    return {
+        "virtual_ip": questions.PromptQuestion(
+            "Virtual IP to use for the Cluster in HA",
+            default_value="",
+            validation_function=validate_virtual_ip,
+        ),
+    }
+
+
+def validate_virtual_ip(value: str) -> str:
+    """We allow passing an empty IP for virtual_ip"""
+    if value == "":
+        return ""
+    try:
+        return ipaddress.ip_address(value).exploded
+    except ValueError as e:
+        raise ValueError(f"{value} is not a valid IP address: {e}")
+
+
 class DeployHAProxyApplicationStep(DeployMachineApplicationStep):
     """Deploy HAProxy application using Terraform"""
 
+    _KEEPALIVED_CONFIG = KEEPALIVED_CONFIG_KEY
+
     def __init__(
         self,
         client: Client,
         manifest: Manifest,
         jhelper: JujuHelper,
         model: str,
+        deployment_preseed: dict[Any, Any] | None = None,
+        accept_defaults: bool = False,
         refresh: bool = False,
     ):
         super().__init__(
@@ -59,10 +90,46 @@ def __init__(
             "Deploying HAProxy",
             refresh,
         )
+        self.preseed = deployment_preseed or {}
+        self.accept_defaults = accept_defaults
 
     def get_application_timeout(self) -> int:
         return HAPROXY_APP_TIMEOUT
 
+    def has_prompts(self) -> bool:
+        return True
+
+    def prompt(self, console: Console | None = None) -> None:
+        variables = questions.load_answers(
+            self.client, self._KEEPALIVED_CONFIG
+        )
+        variables.setdefault("virtual_ip", "")
+
+        # Set defaults
+        self.preseed.setdefault("virtual_ip", "")
+
+        keepalived_config_bank = questions.QuestionBank(
+            questions=keepalived_questions(),
+            console=console,
+            preseed=self.preseed.get("haproxy"),
+            previous_answers=variables,
+            accept_defaults=self.accept_defaults,
+        )
+
+        variables["virtual_ip"] = keepalived_config_bank.virtual_ip.ask()
+
+        LOG.debug(variables)
+        questions.write_answers(
+            self.client, self._KEEPALIVED_CONFIG, variables
+        )
+
+    def extra_tfvars(self) -> dict[str, Any]:
+        variables: dict[str, Any] = questions.load_answers(
+            self.client, self._KEEPALIVED_CONFIG
+        )
+        variables["haproxy_port"] = 80
+        return variables
+
 
 class AddHAProxyUnitsStep(AddMachineUnitsStep):
     """Add HAProxy Unit."""
@@ -114,15 +181,20 @@ def haproxy_install_steps(
     client: Client,
     manifest: Manifest,
     jhelper: JujuHelper,
-    deployment: LocalDeployment,
+    model: str,
     fqdn: str,
+    accept_defaults: bool,
+    preseed: dict[Any, Any],
 ) -> List[BaseStep]:
     return [
         TerraformInitStep(manifest.get_tfhelper("haproxy-plan")),
         DeployHAProxyApplicationStep(
-            client, manifest, jhelper, deployment.infrastructure_model
-        ),
-        AddHAProxyUnitsStep(
-            client, fqdn, jhelper, deployment.infrastructure_model
+            client,
+            manifest,
+            jhelper,
+            model,
+            accept_defaults=accept_defaults,
+            deployment_preseed=preseed,
         ),
+        AddHAProxyUnitsStep(client, fqdn, jhelper, model),
     ]

anvil-python/anvil/jobs/common.py

@@ -25,7 +25,7 @@ class Role(enum.Enum):
     """The role that the current node will play
 
     This determines if the role will be a region node, a rack/agent node,
-    r database node or a haproxy node. The role will help determine which
+    database node, or a haproxy node. The role will help determine which
     particular services need to be configured and installed on the system.
     """
 

anvil-python/anvil/provider/local/commands.py

@@ -269,7 +269,13 @@ def bootstrap(
     if is_haproxy_node:
         plan4.extend(
             haproxy_install_steps(
-                client, manifest_obj, jhelper, deployment, fqdn
+                client,
+                manifest_obj,
+                jhelper,
+                deployment.infrastructure_model,
+                fqdn,
+                accept_defaults,
+                preseed,
             )
         )
     if is_region_node:
@@ -348,6 +354,7 @@ def _print_output(token: str) -> None:
 
 
 @click.command()
+@click.option("-a", "--accept-defaults", help="Accept all defaults.", is_flag=True)
 @click.option("--token", type=str, help="Join token")
 @click.option(
     "--role",
@@ -365,6 +372,7 @@ def join(
     ctx: click.Context,
     token: str,
     roles: List[Role],
+    accept_defaults: bool = False,
 ) -> None:
     """Join node to the cluster.
 
@@ -413,6 +421,7 @@ def join(
     manifest_obj = Manifest.load_latest_from_clusterdb(
         deployment, include_defaults=True
     )
+    preseed = manifest_obj.deployment_config
 
     machine_id = -1
     machine_id_result = get_step_message(plan1_results, AddJujuMachineStep)
@@ -431,7 +440,13 @@ def join(
     if is_haproxy_node:
         plan2.extend(
             haproxy_install_steps(
-                client, manifest_obj, jhelper, deployment, name
+                client,
+                manifest_obj,
+                jhelper,
+                deployment.infrastructure_model,
+                name,
+                accept_defaults,
+                preseed,
             )
         )
     if is_region_node:

anvil-python/anvil/provider/local/deployment.py

@@ -26,6 +26,8 @@
     LocalDeployment as SunbeamLocalDeployment,
 )
 
+from anvil.commands.haproxy import KEEPALIVED_CONFIG_KEY, keepalived_questions
+
 LOG = logging.getLogger(__name__)
 LOCAL_TYPE = "local"
 
@@ -51,5 +53,19 @@ def generate_preseed(self, console: Console) -> str:
             show_questions(bootstrap_bank, section="bootstrap")
         )
 
+        # HAProxy questions
+        try:
+            variables = load_answers(client, KEEPALIVED_CONFIG_KEY)
+        except ClusterServiceUnavailableException:
+            variables = {}
+        keepalived_config_bank = QuestionBank(
+            questions=keepalived_questions(),
+            console=console,
+            previous_answers=variables,
+        )
+        preseed_content.extend(
+            show_questions(keepalived_config_bank, section="haproxy")
+        )
+
         preseed_content_final = "\n".join(preseed_content)
         return preseed_content_final

anvil-python/anvil/versions.py

@@ -18,13 +18,15 @@
 POSTGRESQL_CHANNEL = "14/candidate"
 PGBOUNCER_CHANNEL = "1/candidate"
 HAPROXY_CHANNEL = "latest/stable"
+KEEPALIVED_CHANNEL = "latest/stable"
 
 MACHINE_CHARMS = {
     "maas-region": MAAS_REGION_CHANNEL,
     "maas-agent": MAAS_AGENT_CHANNEL,
     "haproxy": HAPROXY_CHANNEL,
     "postgresql": POSTGRESQL_CHANNEL,
     "pgbouncer": PGBOUNCER_CHANNEL,
+    "keepalived": KEEPALIVED_CHANNEL,
 }
 K8S_CHARMS: dict[str, str] = {}
 
@@ -70,7 +72,12 @@
             "channel": "charm_haproxy_channel",
             "revision": "charm_haproxy_revision",
             "config": "charm_haproxy_config",
-        }
+        },
+        "keepalived": {
+            "channel": "charm_keepalived_channel",
+            "revision": "charm_keepalived_revision",
+            "config": "charm_keepalived_config",
+        },
     }
 }
 

cloud/etc/deploy-haproxy/main.tf

@@ -30,6 +30,10 @@ data "juju_model" "machine_model" {
   name = var.machine_model
 }
 
+locals {
+  virtual_ip = var.virtual_ip != "" ? { virtual_ip = var.virtual_ip } : {}
+}
+
 resource "juju_application" "haproxy" {
   name  = "haproxy"
   model = data.juju_model.machine_model.name
@@ -44,3 +48,37 @@ resource "juju_application" "haproxy" {
 
   config = var.charm_haproxy_config
 }
+
+resource "juju_application" "keepalived" {
+  count = min(length(var.virtual_ip), 1)
+  name  = "keepalived"
+  model = data.juju_model.machine_model.name
+
+  charm {
+    name     = "keepalived"
+    channel  = var.charm_keepalived_channel
+    revision = var.charm_keepalived_revision
+    base     = "ubuntu@22.04"
+  }
+
+  config = merge(
+    { port = var.haproxy_port },
+    local.virtual_ip,
+    var.charm_keepalived_config,
+  )
+}
+
+resource "juju_integration" "maas-region-haproxy" {
+  count = min(length(var.virtual_ip), 1)
+  model = data.juju_model.machine_model.name
+
+  application {
+    name     = juju_application.haproxy.name
+    endpoint = "juju-info"
+  }
+
+  application {
+    name     = juju_application.keepalived[0].name
+    endpoint = "juju-info"
+  }
+}

cloud/etc/deploy-haproxy/variables.tf

@@ -41,3 +41,33 @@ variable "machine_model" {
   description = "Model to deploy to"
   type        = string
 }
+
+variable "charm_keepalived_channel" {
+  description = "Operator channel for Keepalived"
+  type        = string
+  default     = "latest/stable"
+}
+
+variable "charm_keepalived_revision" {
+  description = "Operator channel revision for Keepalived"
+  type        = number
+  default     = null
+}
+
+variable "virtual_ip" {
+  description = "Virtual IP Address for Keepalived"
+  type        = string
+  default     = ""
+}
+
+variable "haproxy_port" {
+  description = "The port that HAProxy listens on"
+  type        = string
+  default     = 80
+}
+
+variable "charm_keepalived_config" {
+  description = "Operator config for Keepalived deployment"
+  type        = map(string)
+  default     = {}
+}