Suppose my cluster got crashed and need to bring up the entire cluster back? Does HA still helps to run the cluster from latest save point?
Regards Bhaskar On Thu, Sep 26, 2019 at 7:44 PM Sean Hester <sean.hes...@bettercloud.com> wrote: > thanks to everyone for all the replies. > > i think the original concern here with "just" relying on the HA option is > that there are some disaster recovery and data center migration use cases > where the continuity of the job managers is difficult to preserve. but > those are admittedly very edgy use cases. i think it's definitely worth > reviewing the SLAs with our site reliability engineers to see how likely it > would be to completely lose all job managers under an HA configuration. > that small a risk might be acceptable/preferable to a one-off solution. > > @Aleksander, would love to learn more about Zookeeper-less HA. i think i > spotted a thread somewhere between Till and someone (perhaps you) about > that. feel free to DM me. > > thanks again to everyone! > > On Thu, Sep 26, 2019 at 7:32 AM Yang Wang <danrtsey...@gmail.com> wrote: > >> Hi, Aleksandar >> >> Savepoint option in standalone job cluster is optional. If you want to >> always recover >> from the latest checkpoint, just as Aleksandar and Yun Tang said you >> could use the >> high-availability configuration. Make sure the cluster-id is not changed, >> i think the job >> could recover both at exceptionally crash and restart by expectation. >> >> @Aleksandar Mastilovic <amastilo...@sightmachine.com>, we are also have >> an zookeeper-less high-availability implementation[1]. >> Maybe we could have some discussion and contribute this useful feature to >> the community. >> >> [1]. >> https://docs.google.com/document/d/1Z-VdJlPPEQoWT1WLm5woM4y0bFep4FrgdJ9ipQuRv8g/edit >> >> Best, >> Yang >> >> Aleksandar Mastilovic <amastilo...@sightmachine.com> 于2019年9月26日周四 >> 上午4:11写道： >> >>> Would you guys (Flink devs) be interested in our solution for >>> zookeeper-less HA? I could ask the managers how they feel about >>> open-sourcing the improvement. >>> >>> On Sep 25, 2019, at 11:49 AM, Yun Tang <myas...@live.com> wrote: >>> >>> As Aleksandar said, k8s with HA configuration could solve your problem. >>> There already have some discussion about how to implement such HA in k8s if >>> we don't have a zookeeper service: FLINK-11105 [1] and FLINK-12884 [2]. >>> Currently, you might only have to choose zookeeper as high-availability >>> service. >>> >>> [1] https://issues.apache.org/jira/browse/FLINK-11105 >>> [2] https://issues.apache.org/jira/browse/FLINK-12884 >>> >>> Best >>> Yun Tang >>> ------------------------------ >>> *From:* Aleksandar Mastilovic <amastilo...@sightmachine.com> >>> *Sent:* Thursday, September 26, 2019 1:57 >>> *To:* Sean Hester <sean.hes...@bettercloud.com> >>> *Cc:* Hao Sun <ha...@zendesk.com>; Yuval Itzchakov <yuva...@gmail.com>; >>> user <user@flink.apache.org> >>> *Subject:* Re: Challenges Deploying Flink With Savepoints On Kubernetes >>> >>> Can’t you simply use JobManager in HA mode? It would pick up where it >>> left off if you don’t provide a Savepoint. >>> >>> On Sep 25, 2019, at 6:07 AM, Sean Hester <sean.hes...@bettercloud.com> >>> wrote: >>> >>> thanks for all replies! i'll definitely take a look at the Flink k8s >>> Operator project. >>> >>> i'll try to restate the issue to clarify. this issue is specific to >>> starting a job from a savepoint in job-cluster mode. in these cases the Job >>> Manager container is configured to run a single Flink job at start-up. the >>> savepoint needs to be provided as an argument to the entrypoint. the Flink >>> documentation for this approach is here: >>> >>> >>> https://github.com/apache/flink/tree/master/flink-container/kubernetes#resuming-from-a-savepoint >>> >>> the issue is that taking this approach means that the job will *always* >>> start >>> from the savepoint provided as the start argument in the Kubernetes YAML. >>> this includes unplanned restarts of the job manager, but we'd really prefer >>> any *unplanned* restarts resume for the most recent checkpoint instead >>> of restarting from the configured savepoint. so in a sense we want the >>> savepoint argument to be transient, only being used during the initial >>> deployment, but this runs counter to the design of Kubernetes which always >>> wants to restore a deployment to the "goal state" as defined in the YAML. >>> >>> i hope this helps. if you want more details please let me know, and >>> thanks again for your time. >>> >>> >>> On Tue, Sep 24, 2019 at 1:09 PM Hao Sun <ha...@zendesk.com> wrote: >>> >>> I think I overlooked it. Good point. I am using Redis to save the path >>> to my savepoint, I might be able to set a TTL to avoid such issue. >>> >>> Hao Sun >>> >>> >>> On Tue, Sep 24, 2019 at 9:54 AM Yuval Itzchakov <yuva...@gmail.com> >>> wrote: >>> >>> Hi Hao, >>> >>> I think he's exactly talking about the usecase where the JM/TM restart >>> and they come back up from the latest savepoint which might be stale by >>> that time. >>> >>> On Tue, 24 Sep 2019, 19:24 Hao Sun, <ha...@zendesk.com> wrote: >>> >>> We always make a savepoint before we shutdown the job-cluster. So the >>> savepoint is always the latest. When we fix a bug or change the job graph, >>> it can resume well. >>> We only use checkpoints for unplanned downtime, e.g. K8S killed JM/TM, >>> uncaught exception, etc. >>> >>> Maybe I do not understand your use case well, I do not see a need to >>> start from checkpoint after a bug fix. >>> From what I know, currently you can use checkpoint as a savepoint as well >>> >>> Hao Sun >>> >>> >>> On Tue, Sep 24, 2019 at 7:48 AM Yuval Itzchakov <yuva...@gmail.com> >>> wrote: >>> >>> AFAIK there's currently nothing implemented to solve this problem, but >>> working on a possible fix can be implemented on top of >>> https://github.com/lyft/flinkk8soperator which already has a pretty >>> fancy state machine for rolling upgrades. I'd love to be involved as this >>> is an issue I've been thinking about as well. >>> >>> Yuval >>> >>> On Tue, Sep 24, 2019 at 5:02 PM Sean Hester <sean.hes...@bettercloud.com> >>> wrote: >>> >>> hi all--we've run into a gap (knowledge? design? tbd?) for our use cases >>> when deploying Flink jobs to start from savepoints using the job-cluster >>> mode in Kubernetes. >>> >>> we're running a ~15 different jobs, all in job-cluster mode, using a mix >>> of Flink 1.8.1 and 1.9.0, under GKE (Google Kubernetes Engine). these are >>> all long-running streaming jobs, all essentially acting as microservices. >>> we're using Helm charts to configure all of our deployments. >>> >>> we have a number of use cases where we want to restart jobs from a >>> savepoint to replay recent events, i.e. when we've enhanced the job logic >>> or fixed a bug. but after the deployment we want to have the job resume >>> it's "long-running" behavior, where any unplanned restarts resume from the >>> latest checkpoint. >>> >>> the issue we run into is that any obvious/standard/idiomatic Kubernetes >>> deployment includes the savepoint argument in the configuration. if the Job >>> Manager container(s) have an unplanned restart, when they come back up they >>> will start from the savepoint instead of resuming from the latest >>> checkpoint. everything is working as configured, but that's not exactly >>> what we want. we want the savepoint argument to be transient somehow (only >>> used during the initial deployment), but Kubernetes doesn't really support >>> the concept of transient configuration. >>> >>> i can see a couple of potential solutions that either involve custom >>> code in the jobs or custom logic in the container (i.e. a custom entrypoint >>> script that records that the configured savepoint has already been used in >>> a file on a persistent volume or GCS, and potentially when/why/by which >>> deployment). but these seem like unexpected and hacky solutions. before we >>> head down that road i wanted to ask: >>> >>> - is this is already a solved problem that i've missed? >>> - is this issue already on the community's radar? >>> >>> thanks in advance! >>> >>> -- >>> *Sean Hester* | Senior Staff Software Engineer | m. 404-828-0865 >>> 3525 Piedmont Rd. NE, Building 6, Suite 500, Atlanta, GA 30305 >>> <http://www.bettercloud.com/> <http://www.bettercloud.com/> >>> *Altitude 2019 in San Francisco | Sept. 23 - 25* >>> It’s not just an IT conference, it’s “a complete learning and networking >>> experience” >>> <https://altitude.bettercloud.com/?utm_source=gmail&utm_medium=signature&utm_campaign=2019-altitude> >>> >>> >>> >>> -- >>> Best Regards, >>> Yuval Itzchakov. >>> >>> >>> >>> -- >>> *Sean Hester* | Senior Staff Software Engineer | m. 404-828-0865 >>> 3525 Piedmont Rd. NE, Building 6, Suite 500, Atlanta, GA 30305 >>> <http://www.bettercloud.com/> <http://www.bettercloud.com/> >>> *Altitude 2019 in San Francisco | Sept. 23 - 25* >>> It’s not just an IT conference, it’s “a complete learning and networking >>> experience” >>> <https://altitude.bettercloud.com/?utm_source=gmail&utm_medium=signature&utm_campaign=2019-altitude> >>> >>> >>> > > -- > *Sean Hester* | Senior Staff Software Engineer | m. 404-828-0865 > 3525 Piedmont Rd. NE, Building 6, Suite 500, Atlanta, GA 30305 > <http://www.bettercloud.com> <http://www.bettercloud.com> > *Altitude 2019 in San Francisco | Sept. 23 - 25* > It’s not just an IT conference, it’s “a complete learning and networking > experience” > <https://altitude.bettercloud.com/?utm_source=gmail&utm_medium=signature&utm_campaign=2019-altitude> > >
