baremetal: keepalived/coredns pods remove before create

If the container exists unexpectedly for any reason, then ExecStop not called, and sometimes there are "storage for container removed" errors trying to re-start it, presumably because podman cleanup removes resources for exited containers.
openshift-metal3 · Jul 18, 2019 · 7c7d13d · 7c7d13d
1 parent 839d40f
commit 7c7d13d
Show file tree

Hide file tree

Showing 2 changed files with 29 additions and 23 deletions.
diff --git a/data/data/bootstrap/baremetal/files/usr/local/bin/coredns.sh b/data/data/bootstrap/baremetal/files/usr/local/bin/coredns.sh
@@ -25,15 +25,18 @@ if ! podman inspect "$COREDNS_IMAGE" &>/dev/null; then
     podman pull "$COREDNS_IMAGE"
 fi
 MATCHES="$(sudo podman ps -a --format "{{.Names}}" | awk '/coredns$/ {print $0}')"
-if [[ -z "$MATCHES" ]]; then
-    /usr/bin/podman create \
-        --name coredns \
-        --volume /etc/coredns:/etc/coredns:z \
-        --network host \
-        --env CLUSTER_DOMAIN="$CLUSTER_DOMAIN" \
-        --env CLUSTER_NAME="$CLUSTER_NAME" \
-        --env NUM_DNS_MEMBERS="$NUM_DNS_MEMBERS" \
-        --env API_VIP="$API_VIP" \
-        "${COREDNS_IMAGE}" \
-            --conf /etc/coredns/Corefile
+if [[ ! -z "$MATCHES" ]]; then
+    # Remove old pod, or you can get storage for container removed
+    # errors on restart if the container exits unexpectedly
+    podman rm -f coredns
 fi
+/usr/bin/podman create \
+    --name coredns \
+    --volume /etc/coredns:/etc/coredns:z \
+    --network host \
+    --env CLUSTER_DOMAIN="$CLUSTER_DOMAIN" \
+    --env CLUSTER_NAME="$CLUSTER_NAME" \
+    --env NUM_DNS_MEMBERS="$NUM_DNS_MEMBERS" \
+    --env API_VIP="$API_VIP" \
+    "${COREDNS_IMAGE}" \
+        --conf /etc/coredns/Corefile
diff --git a/data/data/bootstrap/baremetal/files/usr/local/bin/keepalived.sh b/data/data/bootstrap/baremetal/files/usr/local/bin/keepalived.sh
@@ -40,16 +40,19 @@ export DNS_VRID
 export NET_MASK
 envsubst < /etc/keepalived/keepalived.conf.tmpl | sudo tee /etc/keepalived/keepalived.conf
 
-MATCHES="$(sudo podman ps -a --format "{{.Names}}" | awk '/keepalived$/ {print $0}')"
-if [[ -z "$MATCHES" ]]; then
-    # TODO(bnemec): Figure out how to run with less perms
-    podman create \
-            --name keepalived \
-            --volume /etc/keepalived:/etc/keepalived:z \
-            --network=host \
-            --privileged \
-            --cap-add=ALL \
-            "${KEEPALIVED_IMAGE}" \
-            /usr/sbin/keepalived -f /etc/keepalived/keepalived.conf \
-                --dont-fork -D -l -P
+MATCHES="$(podman ps -a --format "{{.Names}}" | awk '/keepalived$/ {print $0}')"
+if [[ ! -z "$MATCHES" ]]; then
+    # Remove old pod, or you can get storage for container removed
+    # errors on restart if the container exits unexpectedly
+    podman rm -f keepalived
 fi
+# TODO(bnemec): Figure out how to run with less perms
+podman create \
+        --name keepalived \
+        --volume /etc/keepalived:/etc/keepalived:z \
+        --network=host \
+        --privileged \
+        --cap-add=ALL \
+        "${KEEPALIVED_IMAGE}" \
+        /usr/sbin/keepalived -f /etc/keepalived/keepalived.conf \
+            --dont-fork -D -l -P