You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@storm.apache.org by Manikandan <ma...@gmail.com> on 2014/03/14 23:20:17 UTC
Worker Task id during recovery
Hi
When the worker dies and brought back by the supervisor, whats the task id
given to the new worker? Is it always same/different as/from the old one?
Thanks & Regards
Manikandan
Re: Worker Task id during recovery
Posted by Noel Milton Vega <nm...@ComputingArchitects.Com>.
Yes... and I had to figure out the answer to that same question about
two months ago myself. It was a question I
asked myself generally as:
What run-time attributes of a task are remembered/persisted across
task fail & restart sequences?.
Stated that way, the question involved not only task-id and task-index
persistence, but also whether Storm might
redistribute any Field Group that the failed task may have been
receiving, to surviving tasks... after all, the
in-memory state of the failed task was gone, so it's was possible to
redistribute it's Fields Group somewhere
else (in the interest of keeping things moving in Storm).
I couldn't find documentation on this and an understanding was necessary
because, for my purposes, it determined
whether-or-not (and how) I could recover the in-memory state of a failed
task (without replay).
So I took to instrumenting tests using REDIS Sets which, because of item
uniqueness in Sets, allowed me to easily
spot run-time attribute changes of a task before and after simulated
failures (via UNIX 'kill -9').
I wrote up my setup and findings in following little PDF document.
Basically what I found, at least anecdotally,
was that the task-ids, task-indices, as well as their Fields Groups (if
any) were preserved. This, in turn,
meant I could create task-unique identifiers like so, 'fqid =
componentID + . + taskIndex', as unique keys
into REDIS datastructures from which I could recover because,
essentially, 'fqid' didn't change before and after
failure.
Maybe this will help you...
http://didata.us/assets.d/DiData-StormRedis-Portfolio.pdf
Regards,
Noel
On 03/14/2014 06:35 PM, Rick.C.Rankin@kp.org wrote:
> I believe once the topology is submitted you can count on the task
> id's remaining the same. I use this to reload object data if one is
> restarted. They could move to a different slot but should have the
> same id.
>
>
>
> Rick Rankin
> Principal, Analytics & Assurance Services
> Technical Risk Office: Cyber Security
> Kaiser Permanente
> Phone: 510.847.5269
>
>
>
>
> *NOTICE TO RECIPIENT:* If you are not the intended recipient of this
> e-mail, you are prohibited from sharing, copying, or otherwise using
> or disclosing its contents. If you have received this e-mail in
> error, please notify the sender immediately by reply e-mail and
> permanently delete this e-mail and any attachments without reading,
> forwarding or saving them. Thank you.
>
>
>
>
>
> From: Manikandan <ma...@gmail.com>
> To: user@storm.incubator.apache.org
> Date: 03/14/2014 03:21 PM
> Subject: Worker Task id during recovery
> ------------------------------------------------------------------------
>
>
>
> Hi
>
> When the worker dies and brought back by the supervisor, whats the
> task id given to the new worker? Is it always same/different as/from
> the old one?
>
> Thanks & Regards
> Manikandan
Re: Worker Task id during recovery
Posted by Ri...@kp.org.
I believe once the topology is submitted you can count on the task id's
remaining the same. I use this to reload object data if one is restarted.
They could move to a different slot but should have the same id.
Rick Rankin
Principal, Analytics & Assurance Services
Technical Risk Office: Cyber Security
Kaiser Permanente
Phone: 510.847.5269
NOTICE TO RECIPIENT: If you are not the intended recipient of this
e-mail, you are prohibited from sharing, copying, or otherwise using or
disclosing its contents. If you have received this e-mail in error,
please notify the sender immediately by reply e-mail and permanently
delete this e-mail and any attachments without reading, forwarding or
saving them. Thank you.
From: Manikandan <ma...@gmail.com>
To: user@storm.incubator.apache.org
Date: 03/14/2014 03:21 PM
Subject: Worker Task id during recovery
Hi
When the worker dies and brought back by the supervisor, whats the task id
given to the new worker? Is it always same/different as/from the old one?
Thanks & Regards
Manikandan