fix: upgrade K8s client library to add Keep Alive to watches

This has been an outstanding issue for over a month, especially prevalent in AKS. After some inactivity in the cluster, the persistent connection to the K8s API server got interrupted and was never able to recover.

This new fix in the K8s client library puts a 30s keep alive to the Watch connection so that it does not get interrupted.

Author: ivanstanev

Dockerfile

@@ -34,6 +34,9 @@ RUN chmod 755 /usr/bin/dumb-init
 RUN groupadd -g 10001 snyk
 RUN useradd -g snyk -d /srv/app -u 10001 snyk
 
+# @kubernetes/[email protected] started using net-keepalive, which requires the following packages to build modules
+RUN yum --disableplugin=subscription-manager install -y make gcc gcc-c++
+
 WORKDIR /srv/app
 
 COPY --chown=snyk:snyk --from=skopeo-build /usr/bin/skopeo /usr/bin/skopeo
@@ -50,6 +53,8 @@ RUN mkdir -p .config
 
 RUN npm install
 
+RUN yum remove -y make gcc gcc-c++
+
 # add the rest of the app files
 ADD --chown=snyk:snyk . .
 

package-lock.json

@@ -33,7 +33,7 @@
     "lint": "eslint \"src/**/*.ts\" && (cd test && eslint \"**/*.ts\")"
   },
   "dependencies": {
-    "@kubernetes/client-node": "^0.14.0",
+    "@kubernetes/client-node": "^0.14.2",
     "@snyk/dep-graph": "^1.28.0",
     "async": "^3.2.0",
     "aws-sdk": "^2.867.0",