Add REQUEST_TOO_LARGE workflow task failed cause and expose completion size limit (#838)

**What changed?**
Added `WORKFLOW_TASK_FAILED_CAUSE_REQUEST_TOO_LARGE` to the workflow
task failed cause enum. Added
`NamespaceInfo.Limits.workflow_task_completion_size_limit_error` to
communicate the server-enforced max RespondWorkflowTaskCompleted request
size via DescribeNamespace.

**Why?**
When a workflow task completion exceeds the server's size limit we
currently fail with the generic `PAYLOADS_TOO_LARGE`, which is
misleading (no single payload is oversized). The new cause names the
actual condition, and exposing the limit lets SDKs stay under it instead
of discovering it by failing.

**Breaking changes**
None. Additive-only

**Server PR**
TODO

Author: simvlad

openapi/openapiv2.json

@@ -11106,6 +11106,11 @@
           "type": "string",
           "format": "int64",
           "description": "Maximum total memo size in bytes per workflow execution."
+        },
+        "workflowTaskCompletionSizeLimitError": {
+          "type": "string",
+          "format": "int64",
+          "description": "Maximum total size in bytes of a single RespondWorkflowTaskCompleted request.\nRequests exceeding this fail the workflow task with\nWORKFLOW_TASK_FAILED_CAUSE_REQUEST_TOO_LARGE. 0 means no explicit limit."
         }
       }
     },
@@ -22169,10 +22174,11 @@
         "WORKFLOW_TASK_FAILED_CAUSE_GRPC_MESSAGE_TOO_LARGE",
         "WORKFLOW_TASK_FAILED_CAUSE_PAYLOADS_TOO_LARGE",
         "WORKFLOW_TASK_FAILED_CAUSE_EXTERNAL_STORAGE_FAILURE",
-        "WORKFLOW_TASK_FAILED_CAUSE_WORKFLOW_PAUSE_REQUESTED_BEFORE_TASK_STARTED"
+        "WORKFLOW_TASK_FAILED_CAUSE_WORKFLOW_PAUSE_REQUESTED_BEFORE_TASK_STARTED",
+        "WORKFLOW_TASK_FAILED_CAUSE_REQUEST_TOO_LARGE"
       ],
       "default": "WORKFLOW_TASK_FAILED_CAUSE_UNSPECIFIED",
-      "description": "Workflow tasks can fail for various reasons. Note that some of these reasons can only originate\nfrom the server, and some of them can only originate from the SDK/worker.\n\n - WORKFLOW_TASK_FAILED_CAUSE_UNHANDLED_COMMAND: Between starting and completing the workflow task (with a workflow completion command), some\nnew command (like a signal) was processed into workflow history. The outstanding task will be\nfailed with this reason, and a worker must pick up a new task.\n - WORKFLOW_TASK_FAILED_CAUSE_RESET_STICKY_TASK_QUEUE: The worker wishes to fail the task and have the next one be generated on a normal, not sticky\nqueue. Generally workers should prefer to use the explicit `ResetStickyTaskQueue` RPC call.\n - WORKFLOW_TASK_FAILED_CAUSE_NON_DETERMINISTIC_ERROR: The worker encountered a mismatch while replaying history between what was expected, and\nwhat the workflow code actually did.\n - WORKFLOW_TASK_FAILED_CAUSE_PENDING_CHILD_WORKFLOWS_LIMIT_EXCEEDED: We send the below error codes to users when their requests would violate a size constraint\nof their workflow. We do this to ensure that the state of their workflow does not become too\nlarge because that can cause severe performance degradation. You can modify the thresholds for\neach of these errors within your dynamic config.\n\nSpawning a new child workflow would cause this workflow to exceed its limit of pending child\nworkflows.\n - WORKFLOW_TASK_FAILED_CAUSE_PENDING_ACTIVITIES_LIMIT_EXCEEDED: Starting a new activity would cause this workflow to exceed its limit of pending activities\nthat we track.\n - WORKFLOW_TASK_FAILED_CAUSE_PENDING_SIGNALS_LIMIT_EXCEEDED: A workflow has a buffer of signals that have not yet reached their destination. We return this\nerror when sending a new signal would exceed the capacity of this buffer.\n - WORKFLOW_TASK_FAILED_CAUSE_PENDING_REQUEST_CANCEL_LIMIT_EXCEEDED: Similarly, we have a buffer of pending requests to cancel other workflows. We return this error\nwhen our capacity for pending cancel requests is already reached.\n - WORKFLOW_TASK_FAILED_CAUSE_BAD_UPDATE_WORKFLOW_EXECUTION_MESSAGE: Workflow execution update message (update.Acceptance, update.Rejection, or update.Response)\nhas wrong format, or missing required fields.\n - WORKFLOW_TASK_FAILED_CAUSE_UNHANDLED_UPDATE: Similar to WORKFLOW_TASK_FAILED_CAUSE_UNHANDLED_COMMAND, but for updates.\n - WORKFLOW_TASK_FAILED_CAUSE_BAD_SCHEDULE_NEXUS_OPERATION_ATTRIBUTES: A workflow task completed with an invalid ScheduleNexusOperation command.\n - WORKFLOW_TASK_FAILED_CAUSE_PENDING_NEXUS_OPERATIONS_LIMIT_EXCEEDED: A workflow task completed requesting to schedule a Nexus Operation exceeding the server configured limit.\n - WORKFLOW_TASK_FAILED_CAUSE_BAD_REQUEST_CANCEL_NEXUS_OPERATION_ATTRIBUTES: A workflow task completed with an invalid RequestCancelNexusOperation command.\n - WORKFLOW_TASK_FAILED_CAUSE_FEATURE_DISABLED: A workflow task completed requesting a feature that's disabled on the server (either system wide or - typically -\nfor the workflow's namespace).\nCheck the workflow task failure message for more information.\n - WORKFLOW_TASK_FAILED_CAUSE_GRPC_MESSAGE_TOO_LARGE: A workflow task failed because a grpc message was too large.\n - WORKFLOW_TASK_FAILED_CAUSE_PAYLOADS_TOO_LARGE: A workflow task failed because payloads were too large.\n - WORKFLOW_TASK_FAILED_CAUSE_EXTERNAL_STORAGE_FAILURE: A workflow task failed because an external storage operation failed.\nCheck the workflow task failure message for more information.\n - WORKFLOW_TASK_FAILED_CAUSE_WORKFLOW_PAUSE_REQUESTED_BEFORE_TASK_STARTED: A workflow task is failed because the workflow is paused before the task is started."
+      "description": "Workflow tasks can fail for various reasons. Note that some of these reasons can only originate\nfrom the server, and some of them can only originate from the SDK/worker.\n\n - WORKFLOW_TASK_FAILED_CAUSE_UNHANDLED_COMMAND: Between starting and completing the workflow task (with a workflow completion command), some\nnew command (like a signal) was processed into workflow history. The outstanding task will be\nfailed with this reason, and a worker must pick up a new task.\n - WORKFLOW_TASK_FAILED_CAUSE_RESET_STICKY_TASK_QUEUE: The worker wishes to fail the task and have the next one be generated on a normal, not sticky\nqueue. Generally workers should prefer to use the explicit `ResetStickyTaskQueue` RPC call.\n - WORKFLOW_TASK_FAILED_CAUSE_NON_DETERMINISTIC_ERROR: The worker encountered a mismatch while replaying history between what was expected, and\nwhat the workflow code actually did.\n - WORKFLOW_TASK_FAILED_CAUSE_PENDING_CHILD_WORKFLOWS_LIMIT_EXCEEDED: We send the below error codes to users when their requests would violate a size constraint\nof their workflow. We do this to ensure that the state of their workflow does not become too\nlarge because that can cause severe performance degradation. You can modify the thresholds for\neach of these errors within your dynamic config.\n\nSpawning a new child workflow would cause this workflow to exceed its limit of pending child\nworkflows.\n - WORKFLOW_TASK_FAILED_CAUSE_PENDING_ACTIVITIES_LIMIT_EXCEEDED: Starting a new activity would cause this workflow to exceed its limit of pending activities\nthat we track.\n - WORKFLOW_TASK_FAILED_CAUSE_PENDING_SIGNALS_LIMIT_EXCEEDED: A workflow has a buffer of signals that have not yet reached their destination. We return this\nerror when sending a new signal would exceed the capacity of this buffer.\n - WORKFLOW_TASK_FAILED_CAUSE_PENDING_REQUEST_CANCEL_LIMIT_EXCEEDED: Similarly, we have a buffer of pending requests to cancel other workflows. We return this error\nwhen our capacity for pending cancel requests is already reached.\n - WORKFLOW_TASK_FAILED_CAUSE_BAD_UPDATE_WORKFLOW_EXECUTION_MESSAGE: Workflow execution update message (update.Acceptance, update.Rejection, or update.Response)\nhas wrong format, or missing required fields.\n - WORKFLOW_TASK_FAILED_CAUSE_UNHANDLED_UPDATE: Similar to WORKFLOW_TASK_FAILED_CAUSE_UNHANDLED_COMMAND, but for updates.\n - WORKFLOW_TASK_FAILED_CAUSE_BAD_SCHEDULE_NEXUS_OPERATION_ATTRIBUTES: A workflow task completed with an invalid ScheduleNexusOperation command.\n - WORKFLOW_TASK_FAILED_CAUSE_PENDING_NEXUS_OPERATIONS_LIMIT_EXCEEDED: A workflow task completed requesting to schedule a Nexus Operation exceeding the server configured limit.\n - WORKFLOW_TASK_FAILED_CAUSE_BAD_REQUEST_CANCEL_NEXUS_OPERATION_ATTRIBUTES: A workflow task completed with an invalid RequestCancelNexusOperation command.\n - WORKFLOW_TASK_FAILED_CAUSE_FEATURE_DISABLED: A workflow task completed requesting a feature that's disabled on the server (either system wide or - typically -\nfor the workflow's namespace).\nCheck the workflow task failure message for more information.\n - WORKFLOW_TASK_FAILED_CAUSE_GRPC_MESSAGE_TOO_LARGE: A workflow task failed because a grpc message was too large.\n - WORKFLOW_TASK_FAILED_CAUSE_PAYLOADS_TOO_LARGE: A workflow task failed because payloads were too large.\n - WORKFLOW_TASK_FAILED_CAUSE_EXTERNAL_STORAGE_FAILURE: A workflow task failed because an external storage operation failed.\nCheck the workflow task failure message for more information.\n - WORKFLOW_TASK_FAILED_CAUSE_WORKFLOW_PAUSE_REQUESTED_BEFORE_TASK_STARTED: A workflow task is failed because the workflow is paused before the task is started.\n - WORKFLOW_TASK_FAILED_CAUSE_REQUEST_TOO_LARGE: A workflow task failed because the request exceeded a size limit."
     },
     "v1WorkflowTaskFailedEventAttributes": {
       "type": "object",

openapi/openapiv3.yaml

@@ -13451,6 +13451,12 @@ components:
         memoSizeLimitError:
           type: string
           description: Maximum total memo size in bytes per workflow execution.
+        workflowTaskCompletionSizeLimitError:
+          type: string
+          description: |-
+            Maximum total size in bytes of a single RespondWorkflowTaskCompleted request.
+             Requests exceeding this fail the workflow task with
+             WORKFLOW_TASK_FAILED_CAUSE_REQUEST_TOO_LARGE. 0 means no explicit limit.
     NamespaceReplicationConfig:
       type: object
       properties:
@@ -20878,6 +20884,7 @@ components:
             - WORKFLOW_TASK_FAILED_CAUSE_PAYLOADS_TOO_LARGE
             - WORKFLOW_TASK_FAILED_CAUSE_EXTERNAL_STORAGE_FAILURE
             - WORKFLOW_TASK_FAILED_CAUSE_WORKFLOW_PAUSE_REQUESTED_BEFORE_TASK_STARTED
+            - WORKFLOW_TASK_FAILED_CAUSE_REQUEST_TOO_LARGE
           type: string
           format: enum
         failure:

temporal/api/enums/v1/failed_cause.proto

@@ -88,6 +88,8 @@ enum WorkflowTaskFailedCause {
     WORKFLOW_TASK_FAILED_CAUSE_EXTERNAL_STORAGE_FAILURE = 38;
     // A workflow task is failed because the workflow is paused before the task is started.
     WORKFLOW_TASK_FAILED_CAUSE_WORKFLOW_PAUSE_REQUESTED_BEFORE_TASK_STARTED = 39;
+    // A workflow task failed because the request exceeded a size limit.
+    WORKFLOW_TASK_FAILED_CAUSE_REQUEST_TOO_LARGE = 40;
 }
 
 enum StartChildWorkflowExecutionFailedCause {

temporal/api/namespace/v1/message.proto

@@ -77,6 +77,10 @@ message NamespaceInfo {
         int64 blob_size_limit_error = 1;
         // Maximum total memo size in bytes per workflow execution.
         int64 memo_size_limit_error = 2;
+        // Maximum total size in bytes of a single RespondWorkflowTaskCompleted request.
+        // Requests exceeding this fail the workflow task with
+        // WORKFLOW_TASK_FAILED_CAUSE_REQUEST_TOO_LARGE. 0 means no explicit limit.
+        int64 workflow_task_completion_size_limit_error = 3;
     }
 
     // Whether scheduled workflows are supported on this namespace. This is only needed