[AKS] az aks nodepool add/update/upgrade: Add --max-unavailable to specify he maximum number or percentage of nodes that can be simultaneously unavailable during upgrade (#31510)

Author: Jenniferyingni

src/azure-cli/azure/cli/command_modules/acs/_help.py

@@ -1629,6 +1629,9 @@
   - name: --max-surge
     type: string
     short-summary: Extra nodes used to speed upgrade. When specified, it represents the number or percent used, eg. 5 or 33%
+  - name: --max-unavailable
+    type: string
+    short-summary: The maximum number or percentage of nodes that can be simultaneously unavailable during upgrade. When specified, it represents the number or percent used, eg. 1 or 5%
   - name: --drain-timeout
     type: int
     short-summary: When nodes are drain how many minutes to wait for all pods to be evicted
@@ -1790,6 +1793,9 @@
   - name: --max-surge
     type: string
     short-summary: Extra nodes used to speed upgrade. When specified, it represents the number or percent used, eg. 5 or 33%
+  - name: --max-unavailable
+    type: string
+    short-summary: The maximum number or percentage of nodes that can be simultaneously unavailable during upgrade. When specified, it represents the number or percent used, eg. 1 or 5%
   - name: --drain-timeout
     type: int
     short-summary: When nodes are drain how many minutes to wait for all pods to be evicted
@@ -1865,6 +1871,9 @@
   - name: --max-surge
     type: string
     short-summary: Extra nodes used to speed upgrade. When specified, it represents the number or percent used, eg. 5 or 33% (mutually exclusive with "--node-image-only". See "az aks nodepool update --max-surge" to update max surge before upgrading with "--node-image-only")
+  - name: --max-unavailable
+    type: string
+    short-summary: The maximum number or percentage of nodes that can be simultaneously unavailable during upgrade. When specified, it represents the number or percent used, eg. 1 or 5%
   - name: --drain-timeout
     type: int
     short-summary: When nodes are drain how long to wait for all pods to be evicted

src/azure-cli/azure/cli/command_modules/acs/_params.py

@@ -92,7 +92,7 @@
     validate_linux_host_name, validate_load_balancer_idle_timeout,
     validate_load_balancer_outbound_ip_prefixes,
     validate_load_balancer_outbound_ips, validate_load_balancer_outbound_ports,
-    validate_load_balancer_sku, validate_max_surge,
+    validate_load_balancer_sku, validate_max_surge, validate_max_unavailable,
     validate_nat_gateway_idle_timeout,
     validate_nat_gateway_managed_outbound_ip_count, validate_network_policy,
     validate_nodepool_id, validate_nodepool_labels, validate_nodepool_name,
@@ -831,6 +831,7 @@ def load_arguments(self, _):
         c.argument('node_osdisk_type', arg_type=get_enum_type(node_os_disk_types))
         c.argument('node_osdisk_size', type=int)
         c.argument('max_surge', validator=validate_max_surge)
+        c.argument("max_unavailable", validator=validate_max_unavailable)
         c.argument('drain_timeout', type=int)
         c.argument('node_soak_duration', type=int)
         c.argument("undrainable_node_behavior", default='Schedule')
@@ -872,6 +873,7 @@ def load_arguments(self, _):
         c.argument('tags', tags_type)
         c.argument('node_taints', validator=validate_nodepool_taints)
         c.argument('max_surge', validator=validate_max_surge)
+        c.argument("max_unavailable", validator=validate_max_unavailable)
         c.argument('drain_timeout', type=int)
         c.argument('node_soak_duration', type=int)
         c.argument("undrainable_node_behavior")
@@ -891,6 +893,7 @@ def load_arguments(self, _):
 
     with self.argument_context('aks nodepool upgrade') as c:
         c.argument('max_surge', validator=validate_max_surge)
+        c.argument("max_unavailable", validator=validate_max_unavailable)
         c.argument('drain_timeout', type=int)
         c.argument('node_soak_duration', type=int)
         c.argument("undrainable_node_behavior")

src/azure-cli/azure/cli/command_modules/acs/_validators.py

@@ -476,6 +476,21 @@ def validate_max_surge(namespace):
         raise CLIError("--max-surge should be an int or percentage")
 
 
+def validate_max_unavailable(namespace):
+    """validates parameters max unavailable are positive integers or percents."""
+    if namespace.max_unavailable is None:
+        return
+    int_or_percent = namespace.max_unavailable
+    if int_or_percent.endswith('%'):
+        int_or_percent = int_or_percent.rstrip('%')
+
+    try:
+        if int(int_or_percent) < 0:
+            raise InvalidArgumentValueError("--max-unavailable must be positive")
+    except ValueError:
+        raise InvalidArgumentValueError("--max-unavailable should be an int or percentage")
+
+
 def validate_assign_identity(namespace):
     if namespace.assign_identity is not None:
         if namespace.assign_identity == '':

src/azure-cli/azure/cli/command_modules/acs/agentpool_decorator.py

@@ -1103,6 +1103,25 @@ def get_max_surge(self):
         # this parameter does not need validation
         return max_surge
 
+    def get_max_unavailable(self):
+        """Obtain the value of max_unavailable.
+        :return: string
+        """
+        # read the original value passed by the command
+        max_unavailable = self.raw_param.get("max_unavailable")
+        # In create mode, try to read the property value corresponding to the parameter from the `agentpool` object
+        if self.decorator_mode == DecoratorMode.CREATE:
+            if (
+                self.agentpool and
+                self.agentpool.upgrade_settings and
+                self.agentpool.upgrade_settings.max_unavailable is not None
+            ):
+                max_unavailable = self.agentpool.upgrade_settings.max_unavailable
+
+        # this parameter does not need dynamic completion
+        # this parameter does not need validation
+        return max_unavailable
+
     def get_drain_timeout(self):
         """Obtain the value of drain_timeout.
 
@@ -1833,6 +1852,10 @@ def set_up_upgrade_settings(self, agentpool: AgentPool) -> AgentPool:
         if max_surge:
             upgrade_settings.max_surge = max_surge
 
+        max_unavailable = self.context.get_max_unavailable()
+        if max_unavailable:
+            upgrade_settings.max_unavailable = max_unavailable
+
         drain_timeout = self.context.get_drain_timeout()
         if drain_timeout:
             upgrade_settings.drain_timeout_in_minutes = drain_timeout
@@ -2224,6 +2247,10 @@ def update_upgrade_settings(self, agentpool: AgentPool) -> AgentPool:
             # why not always set this? so we don't wipe out a preview feaure in upgrade settigns like NodeSoakDuration?
             agentpool.upgrade_settings = upgrade_settings
 
+        max_unavailable = self.context.get_max_unavailable()
+        if max_unavailable:
+            upgrade_settings.max_unavailable = max_unavailable
+
         drain_timeout = self.context.get_drain_timeout()
         if drain_timeout:
             upgrade_settings.drain_timeout_in_minutes = drain_timeout

src/azure-cli/azure/cli/command_modules/acs/custom.py

@@ -2408,6 +2408,7 @@ def aks_agentpool_add(
     node_osdisk_type=None,
     node_osdisk_size=None,
     max_surge=None,
+    max_unavailable=None,
     drain_timeout=None,
     node_soak_duration=None,
     undrainable_node_behavior=None,
@@ -2477,6 +2478,7 @@ def aks_agentpool_update(
     tags=None,
     node_taints=None,
     max_surge=None,
+    max_unavailable=None,
     drain_timeout=None,
     node_soak_duration=None,
     undrainable_node_behavior=None,
@@ -2530,6 +2532,7 @@ def aks_agentpool_upgrade(cmd, client, resource_group_name, cluster_name,
                           kubernetes_version='',
                           node_image_only=False,
                           max_surge=None,
+                          max_unavailable=None,
                           drain_timeout=None,
                           node_soak_duration=None,
                           undrainable_node_behavior=None,
@@ -2552,11 +2555,12 @@ def aks_agentpool_upgrade(cmd, client, resource_group_name, cluster_name,
         )
 
     # Note: we exclude this option because node image upgrade can't accept nodepool put fields like max surge
-    if (max_surge or drain_timeout or node_soak_duration or undrainable_node_behavior) and node_image_only:
+    hasUpgradeSetting = max_surge or drain_timeout or node_soak_duration or undrainable_node_behavior or max_unavailable
+    if hasUpgradeSetting and node_image_only:
         raise MutuallyExclusiveArgumentError(
-            'Conflicting flags. Unable to specify max-surge/drain-timeout/node-soak-duration with node-image-only.'
+            'Conflicting flags. Unable to specify max-surge/drain-timeout/node-soak-duration/max-unavailable with node-image-only.'
             'If you want to use max-surge/drain-timeout/node-soak-duration with a node image upgrade, please first '
-            'update max-surge/drain-timeout/node-soak-duration using "az aks nodepool update --max-surge/--drain-timeout/--node-soak-duration".'
+            'update max-surge/drain-timeout/node-soak-duration/max-unavailable using "az aks nodepool update --max-surge/--drain-timeout/--node-soak-duration/--max-unavailable".'
         )
 
     if node_image_only: