flink-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Till Rohrmann <trohrm...@apache.org>
Subject Re: Persisting inactive state outside Flink
Date Tue, 18 Feb 2020 10:36:20 GMT
Hi Akshay,

there is no easy out-of-the-box implementation for what you are asking.

Before drafting a potential solution I wanted to ask whether using the
RocksDB state backend could already solve your problem. With this state
backend Flink is able to spill state data to disk. Would this work for your
use case or do you expect the device data per node to grow so big that it
no longer fits onto disk?

If using the RocksDB state backend does not work for you and you really
need to offload state data to an external storage system from where you can
load it lazily it become significantly more complicated. One approach I
could think of is the following: You have a primary operator (process)
which is responsible for processing the incoming events and keeps the state
of the non-dormant devices. Once a device becomes dormant, you could send
the data to a secondary operator (offload+fetching) which uses AsyncIO to
offload the state to Hbase, for example. If the process operator should
encounters an event from a dormant device, it would need to ask the
secondary operator to load it (via sending a fetch event downstream). The
secondary operator would again use AsyncIO to load the requested data. Once
it retrieves the data, you would need to send the data back to the primary
operator via a feedback edge (iteration edge).

The problem with this approach is that Flink does not give you exactly-once
processing guarantees when using iterations at the moment. The community is
working on changing this, though.

Cheers,
Till

On Fri, Feb 14, 2020 at 6:05 PM Akshay Aggarwal <
akshay.aggarwal@flipkart.com> wrote:

> Hi,
>
> We have a use case where we have to persist some state information about a
> device forever. Each new event will fetch the keyed state and update it.
> And this has to be applied in-order of events.
>
> The problem is that the number of devices (keys) will keep growing
> infinitely. Usually a device comes online, stays active for a while
> (generates new events) and then goes into dormant mode. Is there a way we
> can persist the state outside of Flink (say HBase) when the device goes
> dormant and later fetch when it's activated?
>
> I know we can do this in process function using timers. But here I'll have
> to make a synchronous call to the external store every time a new device
> comes live, or when an active device goes dormant, which will stall the
> task and become a scalability bottleneck. Using AsyncIO also doesn't seem
> to be an option.
>
> Is there a way to achieve this without hacking into Flink code?
>
> Thanks,
> Akshay Aggarwal
>
>
> *-----------------------------------------------------------------------------------------*
>
> *This email and any files transmitted with it are confidential and
> intended solely for the use of the individual or entity to whom they are
> addressed. If you have received this email in error, please notify the
> system manager. This message contains confidential information and is
> intended only for the individual named. If you are not the named addressee,
> you should not disseminate, distribute or copy this email. Please notify
> the sender immediately by email if you have received this email by mistake
> and delete this email from your system. If you are not the intended
> recipient, you are notified that disclosing, copying, distributing or
> taking any action in reliance on the contents of this information is
> strictly prohibited.*
>
>
>
> *Any views or opinions presented in this email are solely those of the
> author and do not necessarily represent those of the organization. Any
> information on shares, debentures or similar instruments, recommended
> product pricing, valuations and the like are for information purposes only.
> It is not meant to be an instruction or recommendation, as the case may be,
> to buy or to sell securities, products, services nor an offer to buy or
> sell securities, products or services unless specifically stated to be so
> on behalf of the Flipkart group. Employees of the Flipkart group of
> companies are expressly required not to make defamatory statements and not
> to infringe or authorise any infringement of copyright or any other legal
> right by email communications. Any such communication is contrary to
> organizational policy and outside the scope of the employment of the
> individual concerned. The organization will not accept any liability in
> respect of such communication, and the employee responsible will be
> personally liable for any damages or other liability arising.*
>
>
>
> *Our organization accepts no liability for the content of this email, or
> for the consequences of any actions taken on the basis of the information *
> provided,* unless that information is subsequently confirmed in writing.
> If you are not the intended recipient, you are notified that disclosing,
> copying, distributing or taking any action in reliance on the contents of
> this information is strictly prohibited.*
>
>
> *-----------------------------------------------------------------------------------------*
>
>

Mime
View raw message