Clean up stale keepalived routes in mysql-mgmt init

Add pre-start cleanup in the initcontainer that removes leftover
keepalived routes and VIP addresses before proceeding. Re-deployment
after failure no longer requires manual nsenter cleanup on mgmt nodes.

Closes #36

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Author: szachovy

CHANGELOG.md

@@ -16,6 +16,10 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 * CI workflow to build and push service images to GitHub Container Registry on master merge. (#97)
 * Pull-first with local build fallback for container images during deployment. (#97)
 
+### Fixed
+
+* Clean up stale keepalived routes and VIP addresses before mysql-mgmt init. (#36)
+
 ### Changed
 
 * Completed [ARCHITECTURE.md](./docs/ARCHITECTURE.md) (#93)

services/mysql-mgmt/docker_compose.yml

@@ -12,6 +12,12 @@ services:
       - "-c"
       - |
         set -euo pipefail
+        ip route | grep -E 'proto keepalived|${VIRTUAL_IP_ADDRESS}' | while read -r route; do
+          ip route del $route 2>/dev/null || true
+        done
+        ip addr show dev "${VIRTUAL_NETWORK_INTERFACE}" | grep -oP "${VIRTUAL_IP_ADDRESS}/\d+" | while read -r addr; do
+          ip addr del "$addr" dev "${VIRTUAL_NETWORK_INTERFACE}" 2>/dev/null || true
+        done
         export MYSQL_SUPERSET_PASSWORD=$(</run/secrets/mysql_superset_password)
         if ! mysqlsh \
                --login-path="${PRIMARY_MYSQL_NODE}" \