flink-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Kostas Kloudas <k.klou...@data-artisans.com>
Subject Re: Question about QueryableState
Date Thu, 23 Aug 2018 12:31:36 GMT
Hi Pierre,

You are right that this should not happen.
It seems like a bug.
Could you open a JIRA and post it here?

Thanks,
Kostas

> On Aug 21, 2018, at 9:35 PM, Pierre Zemb <pierre.zemb.isen@gmail.com> wrote:
> 
> Hi!
> 
> I’ve started to deploy a small Flink cluster (4tm and 1jm for now on 1.6.0), and deployed
a small job on it. Because of the current load, job is completely handled by a single tm.
I’ve created a small proxy that is using QueryableStateClient <https://ci.apache.org/projects/flink/flink-docs-master/api/java/org/apache/flink/queryablestate/client/QueryableStateClient.html>
to access the current state. It is working nicely, except under certain circumstances. It
seems to me that I can only access the state through a node that is holding a part of the
job. Here’s an example:
> 
> job on tm1. Pointing QueryableStateClient to tm1. State accessible
> job still on tm1. Pointing QueryableStateClient to tm2 (for example). State inaccessible
> killing tm1, job is now on tm2. State accessible
> job still on tm2. Pointing QueryableStateClient to tm3. State inaccessible
> adding some parallelism to spread job on tm1 and tm2. Pointing QueryableStateClient to
either tm1 and tm2 is working
> job still on tm1 and tm2. Pointing QueryableStateClient to tm3. State inaccessible
> When the state is inaccessible, I can see this (generated here <https://github.com/apache/flink/blob/release-1.6/flink-queryable-state/flink-queryable-state-runtime/src/main/java/org/apache/flink/queryablestate/client/proxy/KvStateClientProxyHandler.java#L228>):
> 
> java.lang.RuntimeException: Failed request 0.
>  Caused by: org.apache.flink.queryablestate.exceptions.UnknownLocationException: Could
not retrieve location of state=repo-status of job=3ac3bc00b2d5bc0752917186a288d40a. Potential
reasons are: i) the state is not ready, or ii) the job does not exist.
>     at org.apache.flink.queryablestate.client.proxy.KvStateClientProxyHandler.getKvStateLookupInfo(KvStateClientProxyHandler.java:228)
>     at org.apache.flink.queryablestate.client.proxy.KvStateClientProxyHandler.getState(KvStateClientProxyHandler.java:162)
>     at org.apache.flink.queryablestate.client.proxy.KvStateClientProxyHandler.executeActionAsync(KvStateClientProxyHandler.java:129)
>     at org.apache.flink.queryablestate.client.proxy.KvStateClientProxyHandler.handleRequest(KvStateClientProxyHandler.java:119)
>     at org.apache.flink.queryablestate.client.proxy.KvStateClientProxyHandler.handleRequest(KvStateClientProxyHandler.java:63)
>     at org.apache.flink.queryablestate.network.AbstractServerHandler$AsyncRequestTask.run(AbstractServerHandler.java:236)
>     at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)
>     at java.util.concurrent.FutureTask.run(FutureTask.java:266)
>     at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
>     at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
>     at java.lang.Thread.run(Thread.java:745)
> From the documentation, I can see that:
> 
> The client connects to a Client Proxy running on a given Task Manager. The proxy is the
entry point of the client to the Flink cluster. It forwards the requests of the client to
the Job Manager and the required Task Manager, and forwards the final response back the client.
> 
> Did I miss something? Is the QueryableStateClientProxy only fetching info from a job
that is running on his local tm? If so, is there a way to retrieve the job-graph? Or maybe
another solution? 
> 
> Thanks!
> Pierre Zemb
> 
> -- 
> Cordialement,
> Pierre Zemb
> pierrezemb.fr <>
> Software Engineer, Metrics Data Platform @OVH


Mime
View raw message