Merge pull request #424 from grac3gao/update1127

TCPXO new release
GoogleCloudPlatform · Dec 16, 2024 · 3140dbc · 3140dbc
2 parents 1ddee13 + 4aeee0d
commit 3140dbc
Show file tree

Hide file tree

Showing 4 changed files with 13 additions and 13 deletions.
diff --git a/gpudirect-tcpxo/nccl-tcpxo-installer.yaml b/gpudirect-tcpxo/nccl-tcpxo-installer.yaml
@@ -69,7 +69,7 @@ spec:
                   chmod a+rw /dev/aperture_devices/*/resource*
               fi
         - name: nccl-tcpxo-installer
-          image: us-docker.pkg.dev/gce-ai-infra/gpudirect-tcpxo/nccl-plugin-gpudirecttcpx-dev:v1.0.6
+          image: us-docker.pkg.dev/gce-ai-infra/gpudirect-tcpxo/nccl-plugin-gpudirecttcpx-dev:v1.0.7
           resources:
             requests:
               cpu: 150m

diff --git a/gpudirect-tcpxo/nccl-test-latest.yaml b/gpudirect-tcpxo/nccl-test-latest.yaml
@@ -80,7 +80,7 @@ spec:
   #  dnsPolicy: ClusterFirstWithHostNet
   containers:
     - name: tcpxo-daemon
-      image: us-docker.pkg.dev/gce-ai-infra/gpudirect-tcpxo/tcpgpudmarxd-dev:v1.0.12
+      image: us-docker.pkg.dev/gce-ai-infra/gpudirect-tcpxo/tcpgpudmarxd-dev:v1.0.13_1
       imagePullPolicy: Always
       command: ["/bin/sh", "-c"]
       args:
@@ -105,7 +105,7 @@ spec:
         - name: LD_LIBRARY_PATH
           value: /usr/local/nvidia/lib64
     - name: nccl-test
-      image: us-docker.pkg.dev/gce-ai-infra/gpudirect-tcpxo/nccl-plugin-gpudirecttcpx-dev:v1.0.6
+      image: us-docker.pkg.dev/gce-ai-infra/gpudirect-tcpxo/nccl-plugin-gpudirecttcpx-dev:v1.0.7
       imagePullPolicy: Always
       #      securityContext:
       #        privileged: true
@@ -210,7 +210,7 @@ spec:
   #  dnsPolicy: ClusterFirstWithHostNet
   containers:
     - name: tcpxo-daemon
-      image: us-docker.pkg.dev/gce-ai-infra/gpudirect-tcpxo/tcpgpudmarxd-dev:v1.0.12
+      image: us-docker.pkg.dev/gce-ai-infra/gpudirect-tcpxo/tcpgpudmarxd-dev:v1.0.13_1
       imagePullPolicy: Always
       command: ["/bin/sh", "-c"]
       args:
@@ -235,7 +235,7 @@ spec:
         - name: LD_LIBRARY_PATH
           value: /usr/local/nvidia/lib64
     - name: nccl-test
-      image: us-docker.pkg.dev/gce-ai-infra/gpudirect-tcpxo/nccl-plugin-gpudirecttcpx-dev:v1.0.6
+      image: us-docker.pkg.dev/gce-ai-infra/gpudirect-tcpxo/nccl-plugin-gpudirecttcpx-dev:v1.0.7
       imagePullPolicy: Always
       #      securityContext:
       #        privileged: true

diff --git a/gpudirect-tcpxo/nccl-test-unprivileged-without-hostnetwork.yaml b/gpudirect-tcpxo/nccl-test-unprivileged-without-hostnetwork.yaml
@@ -66,7 +66,7 @@ spec:
   #  dnsPolicy: ClusterFirstWithHostNet
   containers:
     - name: tcpxo-daemon
-      image: us-docker.pkg.dev/gce-ai-infra/gpudirect-tcpxo/tcpgpudmarxd-dev:v1.0.12
+      image: us-docker.pkg.dev/gce-ai-infra/gpudirect-tcpxo/tcpgpudmarxd-dev:v1.0.13_1
       imagePullPolicy: Always
       command: ["/bin/sh", "-c"]
       args:
@@ -91,7 +91,7 @@ spec:
         - name: LD_LIBRARY_PATH
           value: /usr/local/nvidia/lib64
     - name: nccl-test
-      image: us-docker.pkg.dev/gce-ai-infra/gpudirect-tcpxo/nccl-plugin-gpudirecttcpx-dev:v1.0.6
+      image: us-docker.pkg.dev/gce-ai-infra/gpudirect-tcpxo/nccl-plugin-gpudirecttcpx-dev:v1.0.7
       imagePullPolicy: Always
 #      securityContext:
 #        privileged: true
@@ -196,7 +196,7 @@ spec:
   #  dnsPolicy: ClusterFirstWithHostNet
   containers:
     - name: tcpxo-daemon
-      image: us-docker.pkg.dev/gce-ai-infra/gpudirect-tcpxo/tcpgpudmarxd-dev:v1.0.12
+      image: us-docker.pkg.dev/gce-ai-infra/gpudirect-tcpxo/tcpgpudmarxd-dev:v1.0.13_1
       imagePullPolicy: Always
       command: ["/bin/sh", "-c"]
       args:
@@ -221,7 +221,7 @@ spec:
         - name: LD_LIBRARY_PATH
           value: /usr/local/nvidia/lib64
     - name: nccl-test
-      image: us-docker.pkg.dev/gce-ai-infra/gpudirect-tcpxo/nccl-plugin-gpudirecttcpx-dev:v1.0.6
+      image: us-docker.pkg.dev/gce-ai-infra/gpudirect-tcpxo/nccl-plugin-gpudirecttcpx-dev:v1.0.7
       imagePullPolicy: Always
 #      securityContext:
 #        privileged: true

diff --git a/gpudirect-tcpxo/nccl-test.yaml b/gpudirect-tcpxo/nccl-test.yaml
@@ -41,7 +41,7 @@ spec:
   dnsPolicy: ClusterFirstWithHostNet
   containers:
     - name: tcpxo-daemon
-      image: us-docker.pkg.dev/gce-ai-infra/gpudirect-tcpxo/tcpgpudmarxd-dev:v1.0.12
+      image: us-docker.pkg.dev/gce-ai-infra/gpudirect-tcpxo/tcpgpudmarxd-dev:v1.0.13_1
       imagePullPolicy: Always
       command: ["/bin/sh", "-c"]
       args:
@@ -58,7 +58,7 @@ spec:
         - name: LD_LIBRARY_PATH
           value: /usr/local/nvidia/lib64
     - name: nccl-test
-      image: us-docker.pkg.dev/gce-ai-infra/gpudirect-tcpxo/nccl-plugin-gpudirecttcpx-dev:v1.0.6
+      image: us-docker.pkg.dev/gce-ai-infra/gpudirect-tcpxo/nccl-plugin-gpudirecttcpx-dev:v1.0.7
       imagePullPolicy: Always
       command:
         - /bin/sh
@@ -122,7 +122,7 @@ spec:
   dnsPolicy: ClusterFirstWithHostNet
   containers:
     - name: tcpxo-daemon
-      image: us-docker.pkg.dev/gce-ai-infra/gpudirect-tcpxo/tcpgpudmarxd-dev:v1.0.12
+      image: us-docker.pkg.dev/gce-ai-infra/gpudirect-tcpxo/tcpgpudmarxd-dev:v1.0.13_1
       imagePullPolicy: Always
       command: ["/bin/sh", "-c"]
       args:
@@ -139,7 +139,7 @@ spec:
         - name: LD_LIBRARY_PATH
           value: /usr/local/nvidia/lib64
     - name: nccl-test
-      image: us-docker.pkg.dev/gce-ai-infra/gpudirect-tcpxo/nccl-plugin-gpudirecttcpx-dev:v1.0.6
+      image: us-docker.pkg.dev/gce-ai-infra/gpudirect-tcpxo/nccl-plugin-gpudirecttcpx-dev:v1.0.7
       imagePullPolicy: Always
       command:
         - /bin/sh