Hi Parag,
When you restore from a savepoint do you see a line like: "Restoring job
{} from {}" in jobmanagers logs? Is the entire state lost or just part
of it? Could you explain a bit what does your job look like and how do
you check that the state is lost?Sorry if too obvious, but what are the "accumulators" you refer to? Are they *State primitives[1] or really constructs that are called Accumulator[2]? The latter are not checkpointed. Best, Dawid [1] https://nightlies.apache.org/flink/flink-docs-master/docs/dev/datastream/fault-tolerance/state/#using-keyed-state [2] https://nightlies.apache.org/flink/flink-docs-master/docs/dev/datastream/user_defined_functions/#accumulators--counters On 06/10/2021 08:42, Parag Somani wrote: > Yes Nico. I have evaluated this. > > I have tried below: > > 1. Take the savepoint > 2. Stop the job > 3. Shutdown the instances > 4. Started new pod using below command: > > /docker-entrypoint.sh "standalone-job" "-Ds3.access-key=${AWS_ACCESS_KEY_ID}" > "-Ds3.secret-key=${AWS_SECRET_ACCESS_KEY}" > "-Ds3.endpoint=${AWS_S3_ENDPOINT}" > "-Dhigh-availability.zookeeper.quorum=${ZOOKEEPER_CLUSTER}" "--job-classname" > "com.test.MySpringBootApp" "--fromSavepoint" > "s3://s3-health-service-discovery/savepoints" ${args} > > I haven't observed any errors during start-up in logs. But the state > got reset i.e. values stored inside the accumulator got flushed. > > On Tue, Oct 5, 2021 at 9:40 PM Nicolaus Weidner > <[email protected] > <mailto:[email protected]>> wrote: > > Hi Parag, > > I am not so familiar with the setup you are using, but did you > check out [1]? Maybe the parameter > [--fromSavepoint /path/to/savepoint [--allowNonRestoredState]] > is what you are looking for? > > Best regards, > Nico > > [1] > https://ci.apache.org/projects/flink/flink-docs-master/docs/deployment/resource-providers/standalone/docker/#application-mode-on-docker > > <https://ci.apache.org/projects/flink/flink-docs-master/docs/deployment/resource-providers/standalone/docker/#application-mode-on-docker> > > On Tue, Oct 5, 2021 at 12:37 PM Parag Somani > <[email protected] <mailto:[email protected]>> wrote: > > Hello, > > We are currently using Apache flink 1.12.0 deployed on k8s > cluster of 1.18 with zk for HA. Due to certain vulnerabilities > in container related with few jar(like netty-*, meso), we are > forced to upgrade. > > While upgrading flink to 1.14.0, faced NPE, > > https://issues.apache.org/jira/browse/FLINK-23901?page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel&focusedCommentId=17402570#comment-17402570 > > <https://issues.apache.org/jira/browse/FLINK-23901?page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel&focusedCommentId=17402570#comment-17402570> > > To address it, I have followed steps > > 1. savepoint creation > 2. Stop the job > 3. Restore from save point where i am facing challenge. > > For step #3 from above, i was able to restore from savepoint > mainly because: > "|bin/flink run -s :savepointPath [:runArgs]| " > It majorly about restarting a jar file uploaded. As our > application is based on k8s and running using docker, i was > not able to restore it. And because of it, state of variables > in accumulator got corrupted and i lost the data in one of env. > > My query is, what is preffered way to restore from savepoint, > if application is running on k8s using docker. > > We are using following command to run job manager: > /docker-entrypoint.sh "standalone-job" > "-Ds3.access-key=${AWS_ACCESS_KEY_ID}" > "-Ds3.secret-key=${AWS_SECRET_ACCESS_KEY}" > "-Ds3.endpoint=${AWS_S3_ENDPOINT}" > "-Dhigh-availability.zookeeper.quorum=${ZOOKEEPER_CLUSTER}" "--job-classname" > "<class-name>" ${args} > > Thank you in advance...! > > -- > Regards, > Parag Surajmal Somani. > > > > -- > Regards, > Parag Surajmal Somani.
OpenPGP_signature
Description: OpenPGP digital signature
