flink-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Sampath Bhat <sam414255p...@gmail.com>
Subject Re: Breakage in Flink CLI in 1.5.0
Date Thu, 21 Jun 2018 06:46:31 GMT
Hello Till

Thanks for clarification. But I've few questions based on your reply.

In non-HA setups we need the jobmanager.rpc.address to derive the hostname
of the rest server.
why is there dependency on jobmanager.rpc.address to get the hostname rest
server? This holds good only for normal deployments such as on bare metal,
virtual machine where flink cluster runs as another process in a machine.
But if we try deploy flink on kubernetes then there could be possiblity
where jobmanager.rpc.address and rest.address different from each other.

So if rest.address is not provided in flink-conf.yaml then looking for
jobmanager.rpc.address for deriving the hostname of rest server makes
sense, but when the user has already provided the rest.address but flink
still looks into jobmanager.rpc.address for getting hostname of rest server
is an unwanted dependency IMO.

In HA setup the rpc.address is obtained from zookeeper so user need not
worry about unnecessary properties while submitting job.

On Wed, Jun 20, 2018 at 1:25 PM, Till Rohrmann <trohrmann@apache.org> wrote:

> It will, but it defaults to jobmanager.rpc.address if no rest.address has
> been specified.
>
> On Wed, Jun 20, 2018 at 9:49 AM Chesnay Schepler <chesnay@apache.org>
> wrote:
>
>> Shouldn't the non-HA case be covered by rest.address?
>>
>> On 20.06.2018 09:40, Till Rohrmann wrote:
>>
>> Hi Sampath,
>>
>> it is no longer possible to not start the rest server endpoint by setting
>> rest.port to -1. If you do this, then the cluster won't start. The comment
>> in the flink-conf.yaml holds only true for the legacy mode.
>>
>> In non-HA setups we need the jobmanager.rpc.address to derive the
>> hostname of the rest server. The jobmanager.rpc.port is no longer needed
>> for the client but only for the other cluster components (TMs). When using
>> the HA mode, then every address will be retrieved from ZooKeeper.
>>
>> I hope this clarifies things.
>>
>> Cheers,
>> Till
>>
>> On Wed, Jun 20, 2018 at 9:24 AM Chesnay Schepler <chesnay@apache.org>
>> wrote:
>>
>>> I was worried this might be the case.
>>>
>>> The rest.port handling was simply copied from the legacy web-server,
>>> which explicitly allowed shutting it down.
>>> It may (I'm not entirely sure) also not be necessary for all deployment
>>> modes; for example if the job is baked into the job/taskmanager images.
>>>
>>> I'm not quite sure whether the rpc address is actually required for the
>>> REST job submission, or only since we still rely partly on some legacy
>>> code (ClusterClient). Maybe Till (cc) knows the answer to that.
>>>
>>> > Adding on to this point you made - " the rpc address is still
>>> *required *due
>>> > to some technical implementations; it may be that you can set this to
>>> some
>>> > arbitrary value however."
>>> >
>>> > For job submission to happen successfully we should give specific rpc
>>> > address and not any arbitrary value. If any arbitrary value is given
>>> the
>>> > job submission fails with the following error -
>>> > org.apache.flink.client.deployment.ClusterRetrieveException: Couldn't
>>> > retrieve standalone cluster
>>> >          at
>>> > org.apache.flink.client.deployment.StandaloneClusterDescriptor.
>>> retrieve(StandaloneClusterDescriptor.java:51)
>>> >          at
>>> > org.apache.flink.client.deployment.StandaloneClusterDescriptor.
>>> retrieve(StandaloneClusterDescriptor.java:31)
>>> >          at
>>> > org.apache.flink.client.cli.CliFrontend.runProgram(
>>> CliFrontend.java:249)
>>> >          at org.apache.flink.client.cli.CliFrontend.run(CliFrontend.
>>> java:210)
>>> >          at
>>> > org.apache.flink.client.cli.CliFrontend.parseParameters(
>>> CliFrontend.java:1020)
>>> >          at
>>> > org.apache.flink.client.cli.CliFrontend.lambda$main$9(
>>> CliFrontend.java:1096)
>>> >          at
>>> > org.apache.flink.runtime.security.NoOpSecurityContext.runSecured(
>>> NoOpSecurityContext.java:30)
>>> >          at
>>> > org.apache.flink.client.cli.CliFrontend.main(CliFrontend.java:1096)
>>> > Caused by: java.net.UnknownHostException: flinktest-flink-
>>> jobmanager1233445:
>>> > Name or service not known
>>> >   (Random name flinktest-flink-jobmanager1233445)
>>> >          at java.net.Inet6AddressImpl.lookupAllHostAddr(Native Method)
>>> >          at java.net.InetAddress$2.lookupAllHostAddr(InetAddress.
>>> java:928)
>>> >          at
>>> > java.net.InetAddress.getAddressesFromNameService(
>>> InetAddress.java:1323)
>>> >          at java.net.InetAddress.getAllByName0(InetAddress.java:1276)
>>> >          at java.net.InetAddress.getAllByName(InetAddress.java:1192)
>>> >          at java.net.InetAddress.getAllByName(InetAddress.java:1126)
>>> >          at java.net.InetAddress.getByName(InetAddress.java:1076)
>>> >          at
>>> > org.apache.flink.runtime.rpc.akka.AkkaRpcServiceUtils.
>>> getRpcUrl(AkkaRpcServiceUtils.java:171)
>>> >          at
>>> > org.apache.flink.runtime.rpc.akka.AkkaRpcServiceUtils.
>>> getRpcUrl(AkkaRpcServiceUtils.java:136)
>>> >          at
>>> > org.apache.flink.runtime.highavailability.
>>> HighAvailabilityServicesUtils.createHighAvailabilityServices(
>>> HighAvailabilityServicesUtils.java:83)
>>> >          at
>>> > org.apache.flink.client.program.ClusterClient.<init>(
>>> ClusterClient.java:158)
>>> >          at
>>> > org.apache.flink.client.program.rest.RestClusterClient.<init>(
>>> RestClusterClient.java:184)
>>> >          at
>>> > org.apache.flink.client.program.rest.RestClusterClient.<init>(
>>> RestClusterClient.java:157)
>>> >          at
>>> > org.apache.flink.client.deployment.StandaloneClusterDescriptor.
>>> retrieve(StandaloneClusterDescriptor.java:49)
>>> >          ... 7 more
>>> >
>>> >
>>> > On Wed, Jun 20, 2018 at 11:18 AM, Sampath Bhat <
>>> sam414255path@gmail.com>
>>> > wrote:
>>> >
>>> >> Hi Chesnay
>>> >>
>>> >> If REST API (i.e. the web server) is mandatory for submitting jobs
>>> then
>>> >> why is there an option to set rest.port to -1? I think it should be
>>> >> mandatory to set some valid port for rest.port and make sure flink job
>>> >> manager does not come up if valid port is not set for rest.port? Or
>>> else
>>> >> there must be some way to submit jobs even if REST API (i.e. the web
>>> >> server) is not instantiated.
>>> >>
>>> >> If jobmanger.rpc.address is not required for flink client then why is
>>> it
>>> >> still looking for that property in flink-conf.yaml? Isn't it not a
>>> bug?
>>> >> Because if we comment out the jobmanger.rpc.address and
>>> jobmanger.rpc.port
>>> >> then flink client will not be able to submit the job.
>>> >>
>>> >>
>>> >> On Tue, Jun 19, 2018 at 5:49 PM, Chesnay Schepler <chesnay@apache.org
>>> >
>>> >> wrote:
>>> >>
>>> >>> In 1.5 we reworked the job-submission to go through the REST API
>>> instead
>>> >>> of akka.
>>> >>>
>>> >>> I believe the jobmanager rpc port shouldn't be necessary anymore,
>>> the rpc
>>> >>> address is still *required *due to some technical implementations;
it
>>> >>> may be that you can set this to some arbitrary value however.
>>> >>>
>>> >>> As a result the REST API (i.e. the web server) must be running in
>>> order
>>> >>> to submit jobs.
>>> >>>
>>> >>>
>>> >>> On 19.06.2018 14:12, Sampath Bhat wrote:
>>> >>>
>>> >>> Hello
>>> >>>
>>> >>> I'm using Flink 1.5.0 version and Flink CLI to submit jar to flink
>>> >>> cluster.
>>> >>>
>>> >>> In flink 1.4.2 only job manager rpc address and job manager rpc
port
>>> were
>>> >>> sufficient for flink client to connect to job manager and submit
the
>>> job.
>>> >>>
>>> >>> But in flink 1.5.0 the flink client additionally requires the
>>> >>> rest.address and rest.port for submitting the job to job manager.
>>> What is
>>> >>> the advantage of this new method over the 1.4.2 method of submitting
>>> job?
>>> >>>
>>> >>> Moreover if we make rest.port = -1 the web server will not be
>>> >>> instantiated then how should we submit the job?
>>> >>>
>>> >>> Regards
>>> >>> Sampath
>>> >>>
>>> >>>
>>> >>>
>>>
>>>
>>

Mime
View raw message