WIP: stash changes for blue green aks-preview changes

need to add tests and debug further

Author: CustardTart32

src/aks-preview/azext_aks_preview/_consts.py

@@ -109,6 +109,10 @@
 CONST_NODE_IMAGE_UPGRADE_CHANNEL = "node-image"
 CONST_NONE_UPGRADE_CHANNEL = "none"
 
+# consts for upgrade strategy
+CONST_UPGRADE_STRATEGY_ROLLING = "Rolling"
+CONST_UPGRADE_STRATEGY_BLUE_GREEN = "BlueGreen"
+
 # consts for node os upgrade channel
 CONST_NODE_OS_CHANNEL_NODE_IMAGE = "NodeImage"
 CONST_NODE_OS_CHANNEL_NONE = "None"

src/aks-preview/azext_aks_preview/_help.py

@@ -1879,7 +1879,6 @@
     short-summary: List node pools in the managed Kubernetes cluster.
 """
 
-# TODO: Add params for blue green upgrade settings
 helps['aks nodepool add'] = """
     type: command
     short-summary: Add a node pool to the managed Kubernetes cluster.
@@ -2074,6 +2073,24 @@
         - name: --localdns-config
           type: string
           short-summary: Set the localDNS Profile for a nodepool with a JSON config file.
+        - name: --upgrade-strategy
+          type: string
+          short-summary: Upgrade strategy for the node pool. Allowed values: "Rolling" or "BlueGreen". Default: "Rolling".
+        - name: --drain-batch-size
+          type: string
+          short-summary: Number or percentage of nodes to drain per batch during blue-green upgrades. Accepts an integer (e.g. '5') or percentage (e.g. '50%'). Default: 10%.
+          long-summary: |-
+            Specifies how many nodes to drain in each batch during a blue-green upgrade. Must be a non-zero value, either as an integer (e.g. '5') or a percentage (e.g. '50%') of the total blue nodes at the start of the upgrade. Fractional nodes are rounded up. For more details and best practices, see: https://learn.microsoft.com/en-us/azure/aks/upgrade-cluster
+        - name: --drain-timeout-bg
+          type: int
+          short-summary: Timeout (in minutes) to evict pods and gracefully terminate per node during blue-green upgrades. Default: 30 minutes.
+          long-summary: Maximum time (in minutes) to wait for pod eviction and graceful termination per node during blue-green upgrades. Honors pod disruption budgets. If exceeded, the upgrade fails. Default: 30 minutes.
+        - name: --batch-soak-duration
+          type: int
+          short-summary: Wait time (in minutes) after draining a batch of nodes before proceeding to the next batch. Default: 15 minutes. Only for blue-green upgrades.
+        - name: --final-soak-duration
+          type: int
+          short-summary: Wait time (in minutes) after all old nodes are drained before removing them. Default: 60 minutes. Only for blue-green upgrades.
     examples:
         - name: Create a nodepool in an existing AKS cluster with ephemeral os enabled.
           text: az aks nodepool add -g MyResourceGroup -n nodepool1 --cluster-name MyManagedCluster --node-osdisk-type Ephemeral --node-osdisk-size 48
@@ -2095,6 +2112,8 @@
           text: az aks nodepool add -g MyResourceGroup -n nodepool1 --cluster-name MyManagedCluster --vm-set-type VirtualMachines --vm-sizes "Standard_D4s_v3,Standard_D8s_v3" --node-count 3
         - name: Create a nodepool with ManagedSystem mode
           text: az aks nodepool add -g MyResourceGroup -n managedsystem1 --cluster-name MyManagedCluster --mode ManagedSystem
+        - name: Create a node pool with blue-green upgrade strategy and default parameters
+          text: az aks nodepool add -g MyResourceGroup -n nodepool1 --cluster-name MyManagedCluster --upgrade-strategy BlueGreen
 """
 
 helps['aks nodepool scale'] = """
@@ -2149,6 +2168,24 @@
         - name: --undrainable-node-behavior
           type: string
           short-summary: Define the behavior for undrainable nodes during upgrade. The value should be "Cordon" or "Schedule". The default value is "Schedule".
+        - name: --upgrade-strategy
+          type: string
+          short-summary: Upgrade strategy for the node pool. Allowed values: "Rolling" or "BlueGreen". Default: "Rolling".
+        - name: --drain-batch-size
+          type: string
+          short-summary: Number or percentage of nodes to drain per batch during blue-green upgrades. Accepts an integer (e.g. '5') or percentage (e.g. '50%'). Default: 10%.
+          long-summary: |-
+            Specifies how many nodes to drain in each batch during a blue-green upgrade. Must be a non-zero value, either as an integer (e.g. '5') or a percentage (e.g. '50%') of the total blue nodes at the start of the upgrade. Fractional nodes are rounded up. For more details and best practices, see: https://learn.microsoft.com/en-us/azure/aks/upgrade-cluster
+        - name: --drain-timeout-bg
+          type: int
+          short-summary: Timeout (in minutes) to evict pods and gracefully terminate per node during blue-green upgrades. Default: 30 minutes.
+          long-summary: Maximum time (in minutes) to wait for pod eviction and graceful termination per node during blue-green upgrades. Honors pod disruption budgets. If exceeded, the upgrade fails. Default: 30 minutes.
+        - name: --batch-soak-duration
+          type: int
+          short-summary: Wait time (in minutes) after draining a batch of nodes before proceeding to the next batch. Default: 15 minutes. Only for blue-green upgrades.
+        - name: --final-soak-duration
+          type: int
+          short-summary: Wait time (in minutes) after all old nodes are drained before removing them. Default: 60 minutes. Only for blue-green upgrades.
 """
 
 helps['aks nodepool update'] = """
@@ -2252,6 +2289,24 @@
         - name: --localdns-config
           type: string
           short-summary: Set the localDNS Profile for a nodepool with a JSON config file.
+        - name: --upgrade-strategy
+          type: string
+          short-summary: Upgrade strategy for the node pool. Allowed values: "Rolling" or "BlueGreen". Default: "Rolling".
+        - name: --drain-batch-size
+          type: string
+          short-summary: Number or percentage of nodes to drain per batch during blue-green upgrades. Accepts an integer (e.g. '5') or percentage (e.g. '50%'). Default: 10%.
+          long-summary: |-
+            Specifies how many nodes to drain in each batch during a blue-green upgrade. Must be a non-zero value, either as an integer (e.g. '5') or a percentage (e.g. '50%') of the total blue nodes at the start of the upgrade. Fractional nodes are rounded up. For more details and best practices, see: https://learn.microsoft.com/en-us/azure/aks/upgrade-cluster
+        - name: --drain-timeout-bg
+          type: int
+          short-summary: Timeout (in minutes) to evict pods and gracefully terminate per node during blue-green upgrades. Default: 30 minutes.
+          long-summary: Maximum time (in minutes) to wait for pod eviction and graceful termination per node during blue-green upgrades. Honors pod disruption budgets. If exceeded, the upgrade fails. Default: 30 minutes.
+        - name: --batch-soak-duration
+          type: int
+          short-summary: Wait time (in minutes) after draining a batch of nodes before proceeding to the next batch. Default: 15 minutes. Only for blue-green upgrades.
+        - name: --final-soak-duration
+          type: int
+          short-summary: Wait time (in minutes) after all old nodes are drained before removing them. Default: 60 minutes. Only for blue-green upgrades.
     examples:
       - name: Reconcile the nodepool back to its current state.
         text: az aks nodepool update -g MyResourceGroup -n nodepool1 --cluster-name MyManagedCluster
@@ -2263,6 +2318,8 @@
         text: az aks nodepool update --update-cluster-autoscaler --min-count 1 --max-count 10 -g MyResourceGroup -n nodepool1 --cluster-name MyManagedCluster
       - name: Change a node pool to system mode
         text: az aks nodepool update --mode System -g MyResourceGroup -n nodepool1 --cluster-name MyManagedCluster
+      - name: Update a node pool with blue-green upgrade settings
+        text: az aks nodepool update -g MyResourceGroup -n nodepool1 --cluster-name MyManagedCluster --drain-batch-size 50% --drain-timeout-bg 5 --batch-soak-duration 10 --final-soak-duration 10
 """
 
 helps['aks nodepool get-upgrades'] = """

src/aks-preview/azext_aks_preview/_params.py

@@ -152,6 +152,8 @@
     CONST_ADVANCED_NETWORKPOLICIES_L7,
     CONST_TRANSIT_ENCRYPTION_TYPE_NONE,
     CONST_TRANSIT_ENCRYPTION_TYPE_WIREGUARD,
+    CONST_UPGRADE_STRATEGY_ROLLING,
+    CONST_UPGRADE_STRATEGY_BLUE_GREEN,
 )
 
 from azext_aks_preview._validators import (
@@ -223,6 +225,7 @@
     validate_gateway_prefix_size,
     validate_max_unavailable,
     validate_max_blocked_nodes,
+    validate_drain_batch_size,
     validate_resource_group_parameter,
     validate_location_resource_group_cluster_parameters,
 )
@@ -507,6 +510,11 @@
     CONST_GPU_DRIVER_TYPE_GRID,
 ]
 
+upgrade_strategies = [
+    CONST_UPGRADE_STRATEGY_ROLLING,
+    CONST_UPGRADE_STRATEGY_BLUE_GREEN,
+]
+
 
 def load_arguments(self, _):
     acr_arg_type = CLIArgumentType(metavar="ACR_NAME_OR_RESOURCE_ID")
@@ -1681,7 +1689,12 @@ def load_arguments(self, _):
         c.argument("max_surge", validator=validate_max_surge)
         c.argument("drain_timeout", type=int)
         c.argument("node_soak_duration", type=int)
-        # TODO: Add args for blue green upgrades (and validators if needed)
+        # blue-green upgrade parameters
+        c.argument("upgrade_strategy", arg_type=get_enum_type(upgrade_strategies))
+        c.argument("drain_batch_size", validator=validate_drain_batch_size)
+        c.argument("drain_timeout_bg", type=int)
+        c.argument("batch_soak_duration", type=int)
+        c.argument("final_soak_duration", type=int)
         c.argument("undrainable_node_behavior")
         c.argument("max_unavailable", validator=validate_max_unavailable)
         c.argument("max_blocked_nodes", validator=validate_max_blocked_nodes)
@@ -1821,6 +1834,12 @@ def load_arguments(self, _):
         c.argument("drain_timeout", type=int)
         c.argument("node_soak_duration", type=int)
         c.argument("undrainable_node_behavior")
+        # blue-green upgrade parameters
+        c.argument("upgrade_strategy", arg_type=get_enum_type(upgrade_strategies))
+        c.argument("drain_batch_size", validator=validate_drain_batch_size)
+        c.argument("drain_timeout_bg", type=int)
+        c.argument("batch_soak_duration", type=int)
+        c.argument("final_soak_duration", type=int)
         c.argument("max_unavailable", validator=validate_max_unavailable)
         c.argument("max_blocked_nodes", validator=validate_max_blocked_nodes)
         c.argument("mode", arg_type=get_enum_type(node_mode_types))
@@ -1898,6 +1917,12 @@ def load_arguments(self, _):
         c.argument("drain_timeout", type=int)
         c.argument("node_soak_duration", type=int)
         c.argument("undrainable_node_behavior")
+        # blue-green upgrade parameters
+        c.argument("upgrade_strategy", arg_type=get_enum_type(upgrade_strategies))
+        c.argument("drain_batch_size", validator=validate_drain_batch_size)
+        c.argument("drain_timeout_bg", type=int)
+        c.argument("batch_soak_duration", type=int)
+        c.argument("final_soak_duration", type=int)
         c.argument("max_unavailable", validator=validate_max_unavailable)
         c.argument("max_blocked_nodes", validator=validate_max_blocked_nodes)
         c.argument("snapshot_id", validator=validate_snapshot_id)

src/aks-preview/azext_aks_preview/_validators.py

@@ -530,6 +530,23 @@ def validate_max_blocked_nodes(namespace):
         raise InvalidArgumentValueError('--max-blocked-nodes should be an int or percentage')
 
 
+def validate_drain_batch_size(namespace):
+    """validates drain batch size parameter as non-zero integers or percentages."""
+    if namespace.drain_batch_size is None:
+        return
+    int_or_percent = namespace.drain_batch_size
+    if int_or_percent.endswith('%'):
+        int_or_percent = int_or_percent.rstrip('%')
+
+    try:
+        value = int(int_or_percent)
+        if value <= 0:
+            raise InvalidArgumentValueError('--drain-batch-size must be a non-zero value')
+    except ValueError:
+        # pylint: disable=raise-missing-from
+        raise InvalidArgumentValueError('--drain-batch-size should be an integer or percentage (e.g., "5" or "50%")')
+
+
 def validate_assign_identity(namespace):
     if namespace.assign_identity is not None:
         if namespace.assign_identity == '':