hadoop-mapreduce-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jian He <...@hortonworks.com>
Subject Re: very long timeout on failed RM connect
Date Wed, 12 Feb 2014 22:43:51 GMT
ipc.client.connect.retry.interval sets the underlying ipc retry interval

yarn.resourcemanager.connect.retry-interval.ms sets the upper layer
clientRMProxy retry interval.


Each clientRMProxy retry includes one full round of retries of the
underlying ipc. In each ClientRMProxy retry, the max number of underlying
ipc retry is controlled by "ipc.client.connect.max.retries".

Did you try setting both ?


Jian




On Wed, Feb 12, 2014 at 8:36 AM, John Lilley <john.lilley@redpoint.net>wrote:

>  Setting
>
>     conf.set("yarn.resourcemanager.connect.max-wait.ms", "500");
>
>     conf.set("yarn.resourcemanager.connect.retry-interval.ms", "500");
>
> still results in a wait of around 15 seconds.  Setting this:
>
>    conf.set("ipc.client.connect.max.retries", "2");
>
> Also does not help.  Is there a retry parameter that can be set?
>
> Thanks
>
> John
>
>
>
> *From:* John Lilley [mailto:john.lilley@redpoint.net]
> *Sent:* Monday, February 10, 2014 12:12 PM
> *To:* user@hadoop.apache.org
> *Subject:* RE: very long timeout on failed RM connect
>
>
>
> I tried:
>
> conf.set("yarn.resourcemanager.connect.max-wait.ms", "10000");
>
> conf.set("yarn.resourcemanager.connect.retry-interval.ms", "1000");
>
>
>
> But it has no apparent effect.  Still hangs for a very long time.
>
> john
>
>
>
> *From:* Jian He [mailto:jhe@hortonworks.com <jhe@hortonworks.com>]
> *Sent:* Monday, February 10, 2014 11:05 AM
> *To:* user@hadoop.apache.org
> *Subject:* Re: very long timeout on failed RM connect
>
>
>
> Setting the following two properties may solve your problem.
>
> yarn.resourcemanager.connect.max-wait.ms controls Maximum time to wait to
> establish connection to ResourceManager.
>
> yarn.resourcemanager.connect.retry-interval.ms controls How often to try
> connecting to the ResourceManager.
>
>
>
> Jian
>
>
>
> On Mon, Feb 10, 2014 at 6:44 AM, John Lilley <john.lilley@redpoint.net>
> wrote:
>
> Our application (running outside the Hadoop cluster) connects to the RM
> through YarnClient.  This works fine, except we've found that if the RM
> address or port is misconfigured in our software, or a firewall blocks
> access, the first call into the client (in this case getNodeReports) hangs
> for a very long time.  I've tried
>
>                 conf.set("ipc.client.connect.max.retries", "2");
>
> But this doesn't help.  Is there a configuration setting I can make on the
> YarnClient that will reduce this hang time?
>
> I understand why this long-winded retry strategy exists, in order to
> prevent a highly-loaded cluster from failing jobs.  But it is not
> appropriate for an interactive application.
>
> Thanks
>
> John
>
>
>
>
>
>
> CONFIDENTIALITY NOTICE
> NOTICE: This message is intended for the use of the individual or entity
> to which it is addressed and may contain information that is confidential,
> privileged and exempt from disclosure under applicable law. If the reader
> of this message is not the intended recipient, you are hereby notified that
> any printing, copying, dissemination, distribution, disclosure or
> forwarding of this communication is strictly prohibited. If you have
> received this communication in error, please contact the sender immediately
> and delete it from your system. Thank You.
>

-- 
CONFIDENTIALITY NOTICE
NOTICE: This message is intended for the use of the individual or entity to 
which it is addressed and may contain information that is confidential, 
privileged and exempt from disclosure under applicable law. If the reader 
of this message is not the intended recipient, you are hereby notified that 
any printing, copying, dissemination, distribution, disclosure or 
forwarding of this communication is strictly prohibited. If you have 
received this communication in error, please contact the sender immediately 
and delete it from your system. Thank You.

Mime
View raw message