diff --git a/docs/dyn/bigtableadmin_v2.projects.instances.appProfiles.html b/docs/dyn/bigtableadmin_v2.projects.instances.appProfiles.html
index b1d391e4231..2c674c64e1e 100644
--- a/docs/dyn/bigtableadmin_v2.projects.instances.appProfiles.html
+++ b/docs/dyn/bigtableadmin_v2.projects.instances.appProfiles.html
@@ -114,6 +114,9 @@ <h3>Method Details</h3>
   &quot;description&quot;: &quot;A String&quot;, # Long form description of the use case for this AppProfile.
   &quot;etag&quot;: &quot;A String&quot;, # Strongly validated etag for optimistic concurrency control. Preserve the value returned from `GetAppProfile` when calling `UpdateAppProfile` to fail the request if there has been a modification in the mean time. The `update_mask` of the request need not include `etag` for this protection to apply. See [Wikipedia](https://en.wikipedia.org/wiki/HTTP_ETag) and [RFC 7232](https://tools.ietf.org/html/rfc7232#section-2.3) for more details.
   &quot;multiClusterRoutingUseAny&quot;: { # Read/write requests are routed to the nearest cluster in the instance, and will fail over to the nearest cluster that is available in the event of transient errors or delays. Clusters in a region are considered equidistant. Choosing this option sacrifices read-your-writes consistency to improve availability. # Use a multi-cluster routing policy.
+    &quot;clusterIds&quot;: [ # The set of clusters to route to. The order is ignored; clusters will be tried in order of distance. If left empty, all clusters are eligible.
+      &quot;A String&quot;,
+    ],
   },
   &quot;name&quot;: &quot;A String&quot;, # The unique name of the app profile. Values are of the form `projects/{project}/instances/{instance}/appProfiles/_a-zA-Z0-9*`.
   &quot;singleClusterRouting&quot;: { # Unconditionally routes all read/write requests to a specific cluster. This option preserves read-your-writes consistency but does not improve availability. # Use a single-cluster routing policy.
@@ -136,6 +139,9 @@ <h3>Method Details</h3>
   &quot;description&quot;: &quot;A String&quot;, # Long form description of the use case for this AppProfile.
   &quot;etag&quot;: &quot;A String&quot;, # Strongly validated etag for optimistic concurrency control. Preserve the value returned from `GetAppProfile` when calling `UpdateAppProfile` to fail the request if there has been a modification in the mean time. The `update_mask` of the request need not include `etag` for this protection to apply. See [Wikipedia](https://en.wikipedia.org/wiki/HTTP_ETag) and [RFC 7232](https://tools.ietf.org/html/rfc7232#section-2.3) for more details.
   &quot;multiClusterRoutingUseAny&quot;: { # Read/write requests are routed to the nearest cluster in the instance, and will fail over to the nearest cluster that is available in the event of transient errors or delays. Clusters in a region are considered equidistant. Choosing this option sacrifices read-your-writes consistency to improve availability. # Use a multi-cluster routing policy.
+    &quot;clusterIds&quot;: [ # The set of clusters to route to. The order is ignored; clusters will be tried in order of distance. If left empty, all clusters are eligible.
+      &quot;A String&quot;,
+    ],
   },
   &quot;name&quot;: &quot;A String&quot;, # The unique name of the app profile. Values are of the form `projects/{project}/instances/{instance}/appProfiles/_a-zA-Z0-9*`.
   &quot;singleClusterRouting&quot;: { # Unconditionally routes all read/write requests to a specific cluster. This option preserves read-your-writes consistency but does not improve availability. # Use a single-cluster routing policy.
@@ -182,6 +188,9 @@ <h3>Method Details</h3>
   &quot;description&quot;: &quot;A String&quot;, # Long form description of the use case for this AppProfile.
   &quot;etag&quot;: &quot;A String&quot;, # Strongly validated etag for optimistic concurrency control. Preserve the value returned from `GetAppProfile` when calling `UpdateAppProfile` to fail the request if there has been a modification in the mean time. The `update_mask` of the request need not include `etag` for this protection to apply. See [Wikipedia](https://en.wikipedia.org/wiki/HTTP_ETag) and [RFC 7232](https://tools.ietf.org/html/rfc7232#section-2.3) for more details.
   &quot;multiClusterRoutingUseAny&quot;: { # Read/write requests are routed to the nearest cluster in the instance, and will fail over to the nearest cluster that is available in the event of transient errors or delays. Clusters in a region are considered equidistant. Choosing this option sacrifices read-your-writes consistency to improve availability. # Use a multi-cluster routing policy.
+    &quot;clusterIds&quot;: [ # The set of clusters to route to. The order is ignored; clusters will be tried in order of distance. If left empty, all clusters are eligible.
+      &quot;A String&quot;,
+    ],
   },
   &quot;name&quot;: &quot;A String&quot;, # The unique name of the app profile. Values are of the form `projects/{project}/instances/{instance}/appProfiles/_a-zA-Z0-9*`.
   &quot;singleClusterRouting&quot;: { # Unconditionally routes all read/write requests to a specific cluster. This option preserves read-your-writes consistency but does not improve availability. # Use a single-cluster routing policy.
@@ -213,6 +222,9 @@ <h3>Method Details</h3>
       &quot;description&quot;: &quot;A String&quot;, # Long form description of the use case for this AppProfile.
       &quot;etag&quot;: &quot;A String&quot;, # Strongly validated etag for optimistic concurrency control. Preserve the value returned from `GetAppProfile` when calling `UpdateAppProfile` to fail the request if there has been a modification in the mean time. The `update_mask` of the request need not include `etag` for this protection to apply. See [Wikipedia](https://en.wikipedia.org/wiki/HTTP_ETag) and [RFC 7232](https://tools.ietf.org/html/rfc7232#section-2.3) for more details.
       &quot;multiClusterRoutingUseAny&quot;: { # Read/write requests are routed to the nearest cluster in the instance, and will fail over to the nearest cluster that is available in the event of transient errors or delays. Clusters in a region are considered equidistant. Choosing this option sacrifices read-your-writes consistency to improve availability. # Use a multi-cluster routing policy.
+        &quot;clusterIds&quot;: [ # The set of clusters to route to. The order is ignored; clusters will be tried in order of distance. If left empty, all clusters are eligible.
+          &quot;A String&quot;,
+        ],
       },
       &quot;name&quot;: &quot;A String&quot;, # The unique name of the app profile. Values are of the form `projects/{project}/instances/{instance}/appProfiles/_a-zA-Z0-9*`.
       &quot;singleClusterRouting&quot;: { # Unconditionally routes all read/write requests to a specific cluster. This option preserves read-your-writes consistency but does not improve availability. # Use a single-cluster routing policy.
@@ -255,6 +267,9 @@ <h3>Method Details</h3>
   &quot;description&quot;: &quot;A String&quot;, # Long form description of the use case for this AppProfile.
   &quot;etag&quot;: &quot;A String&quot;, # Strongly validated etag for optimistic concurrency control. Preserve the value returned from `GetAppProfile` when calling `UpdateAppProfile` to fail the request if there has been a modification in the mean time. The `update_mask` of the request need not include `etag` for this protection to apply. See [Wikipedia](https://en.wikipedia.org/wiki/HTTP_ETag) and [RFC 7232](https://tools.ietf.org/html/rfc7232#section-2.3) for more details.
   &quot;multiClusterRoutingUseAny&quot;: { # Read/write requests are routed to the nearest cluster in the instance, and will fail over to the nearest cluster that is available in the event of transient errors or delays. Clusters in a region are considered equidistant. Choosing this option sacrifices read-your-writes consistency to improve availability. # Use a multi-cluster routing policy.
+    &quot;clusterIds&quot;: [ # The set of clusters to route to. The order is ignored; clusters will be tried in order of distance. If left empty, all clusters are eligible.
+      &quot;A String&quot;,
+    ],
   },
   &quot;name&quot;: &quot;A String&quot;, # The unique name of the app profile. Values are of the form `projects/{project}/instances/{instance}/appProfiles/_a-zA-Z0-9*`.
   &quot;singleClusterRouting&quot;: { # Unconditionally routes all read/write requests to a specific cluster. This option preserves read-your-writes consistency but does not improve availability. # Use a single-cluster routing policy.
diff --git a/googleapiclient/discovery_cache/documents/bigtableadmin.v1.json b/googleapiclient/discovery_cache/documents/bigtableadmin.v1.json
index 597677c22f8..22a0f43e19e 100644
--- a/googleapiclient/discovery_cache/documents/bigtableadmin.v1.json
+++ b/googleapiclient/discovery_cache/documents/bigtableadmin.v1.json
@@ -96,7 +96,7 @@
   },
   "protocol": "rest",
   "resources": {},
-  "revision": "20210907",
+  "revision": "20210921",
   "rootUrl": "https://bigtableadmin.googleapis.com/",
   "schemas": {
     "Backup": {
diff --git a/googleapiclient/discovery_cache/documents/bigtableadmin.v2.json b/googleapiclient/discovery_cache/documents/bigtableadmin.v2.json
index 8ffc6fd97a2..ff77d5da554 100644
--- a/googleapiclient/discovery_cache/documents/bigtableadmin.v2.json
+++ b/googleapiclient/discovery_cache/documents/bigtableadmin.v2.json
@@ -1803,7 +1803,7 @@
       }
     }
   },
-  "revision": "20210907",
+  "revision": "20210921",
   "rootUrl": "https://bigtableadmin.googleapis.com/",
   "schemas": {
     "AppProfile": {
@@ -2730,7 +2730,15 @@
     "MultiClusterRoutingUseAny": {
       "description": "Read/write requests are routed to the nearest cluster in the instance, and will fail over to the nearest cluster that is available in the event of transient errors or delays. Clusters in a region are considered equidistant. Choosing this option sacrifices read-your-writes consistency to improve availability.",
       "id": "MultiClusterRoutingUseAny",
-      "properties": {},
+      "properties": {
+        "clusterIds": {
+          "description": "The set of clusters to route to. The order is ignored; clusters will be tried in order of distance. If left empty, all clusters are eligible.",
+          "items": {
+            "type": "string"
+          },
+          "type": "array"
+        }
+      },
       "type": "object"
     },
     "Operation": {