AZP/TEST: Implement remarks by Yossi

openucx · Apr 8, 2024 · 917b2c8 · 917b2c8
1 parent 20ced8f
commit 917b2c8
Show file tree

Hide file tree

Showing 2 changed files with 23 additions and 32 deletions.
diff --git a/buildlib/pr/mad_tests.yml b/buildlib/pr/mad_tests.yml
@@ -21,7 +21,6 @@ jobs:
         displayName: Set Vars
       - bash: |
           source ./buildlib/tools/test_mad.sh
-          node_setup
           build_ucx_in_docker
           docker_run_srv
         displayName: Setup Server
@@ -40,7 +39,6 @@ jobs:
         retryCountOnTaskFailure: 5
       - bash: |
           source ./buildlib/tools/test_mad.sh
-          node_setup
           build_ucx
         displayName: Setup Client
 
@@ -60,9 +58,8 @@ jobs:
       - checkout: none
       - bash: |
           source ./buildlib/tools/test_mad.sh
-          run_mad_test_lid
+          run_mad_test lid $(LID)
         env:
-          LID: $(LID)
           HCA: $(HCA)
         displayName: Test LID
 
@@ -95,9 +92,8 @@ jobs:
       - checkout: none
       - bash: |
           source ./buildlib/tools/test_mad.sh
-          run_mad_test_guid
+          run_mad_test guid $(GUID)
         env:
-          GUID: $(GUID)
           HCA: $(HCA)
         displayName: Test GUID
 

diff --git a/buildlib/tools/test_mad.sh b/buildlib/tools/test_mad.sh
@@ -2,12 +2,12 @@
 set -exE -o pipefail
 
 IMAGE="rdmz-harbor.rdmz.labs.mlnx/ucx/x86_64/rhel8.2/builder:mofed-5.0-1.0.0.0"
-cd "$BUILD_SOURCESDIRECTORY"
 
-node_setup() {
-    funcname
-    sudo chmod 777 /dev/infiniband/umad*
-}
+if [ -z "$BUILD_SOURCESDIRECTORY" ]; then
+    echo "Not running in Azure"
+    exit 1
+fi
+cd "$BUILD_SOURCESDIRECTORY"
 
 build_ucx() {
     funcname
@@ -25,6 +25,7 @@ build_ucx() {
 build_ucx_in_docker() {
     docker run --rm \
         --name ucx_build_"$BUILD_BUILDID" \
+        -e BUILD_SOURCESDIRECTORY="$BUILD_SOURCESDIRECTORY" \
         -v "$PWD":"$PWD" -w "$PWD" \
         -v /hpc/local:/hpc/local \
         $IMAGE \
@@ -34,19 +35,18 @@ build_ucx_in_docker() {
 }
 
 docker_run_srv() {
-    detect_hca
+    HCA=$(detect_hca)
     docker_stop_srv
     docker run --rm \
         --detach \
         --net=host \
-        -e HCA="$HCA" \
         --name ucx_perftest_"$BUILD_BUILDID" \
+        -e BUILD_SOURCESDIRECTORY="$BUILD_SOURCESDIRECTORY" \
         -v "$PWD":"$PWD" -w "$PWD" \
         -v /hpc/local:/hpc/local \
-        --gpus all --ulimit memlock=-1:-1 --device=/dev/infiniband/ \
+        --ulimit memlock=-1:-1 --device=/dev/infiniband/ \
         $IMAGE \
-        bash -c "source ./buildlib/tools/test_mad.sh && \
-                ${PWD}/install/bin/ucx_perftest -K ${HCA}"
+        bash -c "${PWD}/install/bin/ucx_perftest -K ${HCA}"
 }
 
 docker_stop_srv() {
@@ -55,29 +55,25 @@ docker_stop_srv() {
 
 set_vars() {
     set +x
-    detect_hca
+    HCA=$(detect_hca)
     # Replace ':' with space for 'ibstat' format
-    HCA_IBSTAT=${HCA/:/ }
+    HCA_DEV=${HCA/:/ }
     # shellcheck disable=SC2086
-    LID=$(ibstat $HCA_IBSTAT | grep Base | awk '{print $NF}')
+    LID=$(ibstat $HCA_DEV | grep Base | awk '{print $NF}')
     # shellcheck disable=SC2086
-    GUID=$(ibstat $HCA_IBSTAT | grep GUID | awk '{print $NF}')
+    GUID=$(ibstat $HCA_DEV | grep GUID | awk '{print $NF}')
     echo "##vso[task.setvariable variable=LID;isOutput=true]$LID"
     echo "##vso[task.setvariable variable=GUID;isOutput=true]$GUID"
-    echo "##vso[task.setvariable variable=HCA;isOutput=true]$HCA"    
+    echo "##vso[task.setvariable variable=HCA;isOutput=true]$HCA"
     echo "LID: $LID"
     echo "GUID: $GUID"
     echo "HCA: $HCA"
 }
 
-run_mad_test_lid() {
-    funcname
-    "$PWD"/install/bin/ucx_perftest -t tag_bw -e -K "$HCA" -e lid:"$LID"
-}
-
-run_mad_test_guid() {
-    funcname
-    "$PWD"/install/bin/ucx_perftest -t tag_bw -e -K "$HCA" guid:"$GUID"
+run_mad_test() {
+    test_type="$1"
+    ib_add="$2"
+    "$PWD"/install/bin/ucx_perftest -t tag_bw -e -K "$HCA" -e "$test_type":"$ib_add"
 }
 
 funcname() {
@@ -87,7 +83,6 @@ funcname() {
 }
 
 detect_hca() {
-    echo "Detect first active HCA port"
-    HCA="$(ibv_devinfo | awk '/hca_id:/ {hca=$2} /port:/ {port=$2} /PORT_ACTIVE/ {print hca ":" port; exit}')"
-    export HCA
+    # Detect first active HCA port
+    ibv_devinfo | awk '/hca_id:/ {hca=$2} /port:/ {port=$2} /PORT_ACTIVE/ {print hca ":" port; exit}'
 }