bug(al2023): reset CNI cache on boot (#1912)

awslabs · Aug 9, 2024 · d1d2163 · d1d2163
1 parent 6811189
commit d1d2163
Show file tree

Hide file tree

Showing 2 changed files with 15 additions and 0 deletions.
diff --git a/templates/al2023/provisioners/install-worker.sh b/templates/al2023/provisioners/install-worker.sh
@@ -239,3 +239,8 @@ sudo chown -R root:root /etc/eks
 sudo sed -i \
   's/ - package-update-upgrade-install/# Removed so that nodes do not have version skew based on when the node was started.\n# - package-update-upgrade-install/' \
   /etc/cloud/cloud.cfg
+
+# the CNI results cache is not valid across reboots, and errant files can prevent cleanup of pod sandboxes
+# https://github.com/containerd/containerd/issues/8197
+# this was fixed in 1.2.x of libcni but containerd < 2.x are using libcni 1.1.x
+sudo systemctl enable cni-cache-reset
diff --git a/templates/al2023/runtime/rootfs/etc/systemd/system/cni-cache-reset.service b/templates/al2023/runtime/rootfs/etc/systemd/system/cni-cache-reset.service
@@ -0,0 +1,10 @@
+[Unit]
+Description=Reset the CNI results cache
+Before=cloud-init.target
+
+[Service]
+Type=oneshot
+ExecStart=rm -rf /var/lib/cni/results
+
+[Install]
+WantedBy=multi-user.target