From 38f6271773a2682730f2a3aaa02dc9d5a676f703 Mon Sep 17 00:00:00 2001
From: NachoEchevarria <ignacio.echevarria@datadoghq.com>
Date: Mon, 13 Apr 2026 17:38:16 +0200
Subject: [PATCH] Docker run retry

---
 .azure-pipelines/steps/run-in-docker.yml | 24 ++++++++++++++++++++++++
 1 file changed, 24 insertions(+)

diff --git a/.azure-pipelines/steps/run-in-docker.yml b/.azure-pipelines/steps/run-in-docker.yml
index 11b818ca7f51..3f7f5916e981 100644
--- a/.azure-pipelines/steps/run-in-docker.yml
+++ b/.azure-pipelines/steps/run-in-docker.yml
@@ -61,6 +61,30 @@ steps:
 
     echo "Using SDK version $sdkVersion"
 
+    # Retry wrapper for transient Docker daemon errors (e.g., D-Bus/cgroup flakes).
+    # Exit code 125 = Docker daemon error (container creation failure, not a test/build failure).
+    docker_run_with_retry() {
+      local max_retries=3
+      local retry_delay=10
+      local attempt=1
+      while [ "$attempt" -le "$max_retries" ]; do
+        "$@"
+        local exit_code=$?
+        if [ "$exit_code" -eq 0 ]; then
+          return 0
+        fi
+        echo "##[warning][docker-retry] docker run failed with exit code ${exit_code} on attempt ${attempt}/${max_retries}."
+        if [ "$exit_code" -eq 125 ] && [ "$attempt" -lt "$max_retries" ]; then
+          echo "##[warning][docker-retry] Exit code 125 indicates a Docker daemon error. Retrying in ${retry_delay}s..."
+          sleep "$retry_delay"
+          attempt=$((attempt + 1))
+          continue
+        fi
+        return "$exit_code"
+      done
+    }
+
+    docker_run_with_retry \
     docker run --rm \
         --cap-add=SYS_PTRACE \
         --mount type=bind,source="$(System.DefaultWorkingDirectory)",target=/project \