The validity of that advice depends on a lot of factors. G1 changed the
game a bit for pause times for sure but you can still see larger pause
times than acceptable for some cases. In any event I agree that we should
be more careful with how we describe heap usage.
Thanks
Joe
On Oct 14, 2016
We actually use heap sizes from 32 to 64Gb for ours but our volumes and graphs
are both extremely large. Although I believe the smaller heap sizes were a
limitation of the garbage collection in Java 7. We also moved to ssd drives,
which did help through put quite a bit. Our systems were
Thanks James. I would be happy to contribute the scan processor for
DynamoDB. Just to clarify, based on your comment, we can't take all the
rows of the DynamoDB table and put it into another table. We have to do it
for one record at a time?
On Fri, Oct 14, 2016 at 10:50 AM, James Wing
NiFi's GetDynamoDB processor uses the underlying BatchGetItem API, which
requires item keys as inputs. Iterating over the keys in a table would
require the Scan API, but NiFi does not have a processor to scan a DynamoDB
table.
This would be a great addition to NiFi. If you have any interest in
Manish,
You could use ExecuteScript with Groovy and the following:
def flowFile = session.get()
if(!flowFile) return
flowFile = session.penalize(flowFile)
session.transfer(flowFile, REL_SUCCESS)
Then you can route "success" back to the processor. You'd set the
Penalty Duration on the
Thanks Matt. As a workaround, is there a processor that does not modify the
flow file (content or attribute) at all, and I can use it to delay the
self-referencing flow files to hit the main processor again immediately?
Regards,
Manish
-Original Message-
From: Matt Burgess
Manish,
The use of penalize(), yield(), etc. is not enforced by the framework,
so processors can have different behavior, sometimes on purpose, and
sometimes inadvertently. The Developer's Guide has guidance on when
to use such methods [1], and reviewers often check the submissions to
see if
Ali,
"not recommended to dedicate more than 8-10 GM to JVM heap space" by
whom? Do you have links/references establishing this? I couldn't find
anyone saying this or why.
Russ
On 10/13/2016 05:47 PM, Ali Nazemian wrote:
Hi,
I have another question regarding the hardware recommendation. As
Hello Everyone,
In some of the processors I have seen that flow files on failure are not being
penalized. For example - Kite processors like ConvertJsonToAvro. Is there some
specific reason why some processors have different behavior?
I think every processor should penalize every non-success
I'd also add to Mark's great reply that another good use of RAM beyond the
HEAP and disk caching and avoiding swapping is that you can do things like
off-heap native storage of things like reference datasets that can wired
into NiFi flows for high speed enrichment where you can even do hot
Thanks Jeremy. Where can I set/change this property? in GetKafka?
On Thu, Oct 13, 2016 at 5:33 PM, Jeremy Farbota wrote:
> Igor,
>
> Kafka consumer properties can be found here: http://kafka.apache.org/
> documentation.html#consumerconfigs
>
> GetKafka uses the old consumer
Hi Ali,
Typically, we see people using a 4-8 GB heap with NiFi. 8 GB is pretty typical
for a flow that is expected to have
pretty high throughput in terms of the number of FlowFiles, or a large number
of processors. However, one thing
that you will want to consider in terms of RAM is disk
12 matches
Mail list logo