I think I understand.
Since the entrie session must fit into memory anyway I'm going to try to
create a new datastructure which simply contains the 'entire session' and
simply use a Window/BucketingSink construct to ship them to files.
I do need to ensure noone can OOM the system by capping the ma
You are right. Building a large record might result in an OOME.
But I think, that would also happen with a regular SessionWindow,
RocksDBStatebackend, and a WindowFunction that immediately ships the
records it receives from the Iterable.
As far as I know, a SessionWindow stores all elements in an i
Hi Fabian,
On Fri, Jun 30, 2017 at 6:27 PM, Fabian Hueske wrote:
> If I understand your use case correctly, you'd like to hold back all
> events of a session until it ends/timesout and then write all events out.
> So, instead of aggregating per session (the common use case), you'd just
> like to
Hi Niels,
If I understand your use case correctly, you'd like to hold back all events
of a session until it ends/timesout and then write all events out.
So, instead of aggregating per session (the common use case), you'd just
like to collect the event.
I would implement a simple WindowFunction th
Looks like you are missing a window *function* that processes the window.
>From [1] :
stream
.keyBy(...) <- keyed versus non-keyed windows
.window(...) <- required: "assigner"
[.trigger(...)] <- optional: "trigger" (else default trigger)
[.evicto
Hi,
I have the following situation:
- I have click stream data from a website that contains a session id.
- I then use something like EventTimeSessionWindows to group these events
per session into a WindowedStream
So I essentially end up with a stream of "finished sessions"
So far I am able to d
