Merge branch 'main' into sumanth/validate-upload-result

Azure · Oct 15, 2024 · 226b0c5 · 226b0c5
2 parents 5ff2271 + 556ea25
commit 226b0c5
Show file tree

Hide file tree

Showing 4 changed files with 32 additions and 6 deletions.
diff --git a/modules/terraform/aws/virtual-network/main.tf b/modules/terraform/aws/virtual-network/main.tf
@@ -32,6 +32,32 @@ resource "aws_subnet" "subnets" {
   tags = merge(local.tags, {
     "Name" = each.value.name
   })
+
+  provisioner "local-exec" {
+    /*
+      This command cleans up any not in use ENIs attached to this subnet, which were created outside the scope of the Terraform modules.
+      This is a workaround for the known VPC CNI addon's "leaked ENIs" issue: See https://github.com/aws/amazon-vpc-cni-k8s/issues/608
+    */
+    when    = destroy
+    command = <<-EOT
+      echo "Detaching Subnet: ${self.id} Network Interfaces"
+      # Get available (not in use) ENIs attached to this subnet
+      network_interfaces_attachment_ids=$(aws ec2 describe-network-interfaces --filters Name=subnet-id,Values=${self.id} --query "NetworkInterfaces[?Status=='available' && Attachment.AttachIndex != '0'].Attachment.AttachmentId" --output text)
+      for network_interface_attachment_id in $network_interfaces_attachment_ids; do
+        echo "Detaching available Network Interface attachment id: $network_interface_attachment_id"
+        if ! aws ec2 detach-network-interface --attachment-id $network_interface_attachment_id; then
+          echo "##[warning] Failed to detach Network Interface attachment id: $network_interface_attachment_id"
+        fi
+      done
+      network_interfaces=$(aws ec2 describe-network-interfaces --filters Name=subnet-id,Values=${self.id} --query "NetworkInterfaces[?Status=='available' && Attachment.AttachIndex != '0'].NetworkInterfaceId" --output text)
+      for network_interface in $network_interfaces; do
+        echo "Deleting available Network Interface: $network_interface"
+        if ! aws ec2 delete-network-interface --network-interface-id $network_interface; then
+          echo "##[warning] Failed to delete Network Interface: $network_interface"
+        fi
+      done
+    EOT
+  }
 }
 
 resource "aws_eip" "eips" {

diff --git a/pipelines/perf-eval/apiserver-benchmark-virtualnodes10-pods100.yml b/pipelines/perf-eval/apiserver-benchmark-virtualnodes10-pods100.yml
@@ -33,7 +33,7 @@ stages:
               extra_benchmark_subcmd_args: ""
               disable_warmup: "true"
           engine_input:
-            runner_image: telescope.azurecr.io/oss/kperf:v0.1.3
+            runner_image: telescope.azurecr.io/oss/kperf:v0.1.5
             benchmark_subcmd: node10_job1_pod100
             benchmark_subcmd_args: "--total 1000"
           max_parallel: 2
@@ -58,7 +58,7 @@ stages:
               flowcontrol: "exempt:5"
               extra_benchmark_subcmd_args: ""
           engine_input:
-            runner_image: telescope.azurecr.io/oss/kperf:v0.1.3
+            runner_image: telescope.azurecr.io/oss/kperf:v0.1.5
             benchmark_subcmd: node10_job1_pod100
             benchmark_subcmd_args: "--total 1000"
           max_parallel: 2

diff --git a/pipelines/perf-eval/apiserver-benchmark-virtualnodes100-pods10k.yml b/pipelines/perf-eval/apiserver-benchmark-virtualnodes100-pods10k.yml
@@ -37,7 +37,7 @@ stages:
               flowcontrol: "exempt:5"
               extra_benchmark_subcmd_args: "--padding-bytes=16384"
           engine_input:
-            runner_image: telescope.azurecr.io/oss/kperf:v0.1.2
+            runner_image: telescope.azurecr.io/oss/kperf:v0.1.5
             benchmark_subcmd: node100_pod10k
             benchmark_subcmd_args: "--total 72000 --deployments=10 --interval 24h --cpu 64 --memory 192"
           max_parallel: 2
@@ -68,7 +68,7 @@ stages:
               flowcontrol: "exempt:5"
               extra_benchmark_subcmd_args: "--padding-bytes=16384"
           engine_input:
-            runner_image: telescope.azurecr.io/oss/kperf:v0.1.2
+            runner_image: telescope.azurecr.io/oss/kperf:v0.1.5
             benchmark_subcmd: node100_pod10k
             benchmark_subcmd_args: "--total 72000 --deployments=10 --interval 24h --cpu 64 --memory 192"
           max_parallel: 2

diff --git a/pipelines/perf-eval/apiserver-benchmark-virtualnodes100-pods3k.yml b/pipelines/perf-eval/apiserver-benchmark-virtualnodes100-pods3k.yml
@@ -31,7 +31,7 @@ stages:
               flowcontrol: "exempt:5"
               extra_benchmark_subcmd_args: ""
           engine_input:
-            runner_image: telescope.azurecr.io/oss/kperf:v0.0.8
+            runner_image: telescope.azurecr.io/oss/kperf:v0.1.5
             benchmark_subcmd: node100_job1_pod3k
             benchmark_subcmd_args: "--total 36000"
           max_parallel: 2
@@ -56,7 +56,7 @@ stages:
               flowcontrol: "exempt:5"
               extra_benchmark_subcmd_args: ""
           engine_input:
-            runner_image: telescope.azurecr.io/oss/kperf:v0.0.8
+            runner_image: telescope.azurecr.io/oss/kperf:v0.1.5
             benchmark_subcmd: node100_job1_pod3k
             benchmark_subcmd_args: "--total 36000"
           max_parallel: 2