Hi Lilli! I am not aware of any problems with savepoint restore in application mode. What you can try is to use the *execution.savepoint.path *configuration setting to control it instead of the CLI and see if it makes a difference for you.
Otherwise, you could also check out the https://github.com/apache/flink-kubernetes-operator (docs <https://nightlies.apache.org/flink/flink-kubernetes-operator-docs-release-0.1/>) which can help you manage your Flink Application Deployments in Kubernetes. Cheers, Gyula On Mon, Apr 11, 2022 at 8:09 PM Lilli Pearson <epear...@wayfair.com> wrote: > Hi, > > Summary: > I've run into a number of issues trying to marry savepoints with running > Flink in Application Mode, and am wondering if anyone has suggestions on > how to resolve them, or if savepoints and Application Mode simply aren't > designed to work together. > > Context on app deployment: > For long-running processing of my Kafka streams, I'm running Flink 1.13.5 > in application mode, using CI/CD to deploy the cluster to Kubernetes by > deleting and recreating the deployment. This approach has worked great with > checkpoints. However, since the savepoint Flink should start up with needs > to be specified on startup, this approach would need to change a bit. > > Details: > In experimenting with savepoints while running the app in Application > Mode, I've run into some issues that have made me suspect these two > features just don't work well together, at least in Flink 1.13, though I > can't find documentation that says so directly. (Maybe it's implied, as > considering how application mode is set up, it does seem reasonable to me > that savepoints wouldn't work.) For example: > * The entire /jars API (link: > https://nightlies.apache.org/flink/flink-docs-release-1.13/docs/ops/rest_api/#jars) > is simply unavailable on my cluster (404s), though the rest of the API > works fine. This means I can't use those endpoints to submit a jar to start > from > * using the CLI to run has been equally unsuccessful; when I run a command > like `bin/flink run path/to/jar.jar -s path/to/savepoint`, it fails and > the root cause error is > `org.apache.flink.runtime.rest.util.RestClientException: [Not found.]` > > > Thanks in advance for any help or advice! > >
