Allow MapRun redriving (#304)

Fixes #303
NASA-IMPACT · Nov 29, 2023 · e3dab53 · e3dab53
1 parent b77b353
commit e3dab53
Show file tree

Hide file tree

Showing 2 changed files with 17 additions and 2 deletions.
diff --git a/app/stacks/cumulus/iam.tf b/app/stacks/cumulus/iam.tf
@@ -33,6 +33,7 @@ data "aws_iam_policy_document" "allow_sfn_distributed_maps" {
       "states:DescribeExecution",
       "states:StartExecution",
       "states:StopExecution",
+      "states:RedriveExecution",
     ]
     resources = ["*"]
   }

diff --git a/app/stacks/cumulus/templates/discover-granules-workflow.asl.json b/app/stacks/cumulus/templates/discover-granules-workflow.asl.json
@@ -31,7 +31,7 @@
       "Type": "Map",
       "End": true,
       "MaxConcurrency": 10,
-      "ToleratedFailurePercentage": 0,
+      "ToleratedFailurePercentage": 3,
       "ItemReader": {
         "Resource": "arn:aws:states:::s3:getObject",
         "ReaderConfig": {
@@ -89,7 +89,8 @@
                           "collection": "{$.meta.collection}",
                           "buckets": "{$.meta.buckets}",
                           "stack": "{$.meta.stack}",
-                          "duplicateGranuleHandling": "{$.meta.collection.duplicateHandling}"
+                          "duplicateGranuleHandling": "{$.meta.collection.duplicateHandling}",
+                          "concurrency": 8
                         }
                       }
                     },
@@ -98,13 +99,22 @@
                     "Retry": [
                       {
                         "ErrorEquals": [
+                          "Lambda.ClientExecutionTimeoutException",
                           "Lambda.ServiceException",
                           "Lambda.AWSLambdaException",
                           "Lambda.SdkClientException"
                         ],
                         "IntervalSeconds": 2,
                         "MaxAttempts": 6,
                         "BackoffRate": 2
+                      },
+                      {
+                        "ErrorEquals": [
+                          "Lambda.Unknown",
+                          "States.Timeout"
+                        ],
+                        "IntervalSeconds": 30,
+                        "MaxAttempts": 1
                       }
                     ],
                     "Next": "BatchGranules"
@@ -131,6 +141,7 @@
                     "Retry": [
                       {
                         "ErrorEquals": [
+                          "Lambda.ClientExecutionTimeoutException",
                           "Lambda.ServiceException",
                           "Lambda.AWSLambdaException",
                           "Lambda.SdkClientException"
@@ -181,6 +192,7 @@
                           "Retry": [
                             {
                               "ErrorEquals": [
+                                "Lambda.ClientExecutionTimeoutException",
                                 "Lambda.ServiceException",
                                 "Lambda.AWSLambdaException",
                                 "Lambda.SdkClientException"
@@ -230,6 +242,7 @@
                           "Retry": [
                             {
                               "ErrorEquals": [
+                                "Lambda.ClientExecutionTimeoutException",
                                 "Lambda.ServiceException",
                                 "Lambda.AWSLambdaException",
                                 "Lambda.SdkClientException"
@@ -290,6 +303,7 @@
                           "Retry": [
                             {
                               "ErrorEquals": [
+                                "Lambda.ClientExecutionTimeoutException",
                                 "Lambda.ServiceException",
                                 "Lambda.AWSLambdaException",
                                 "Lambda.SdkClientException"