2019-07-21 17:55:38 UTC - mapa: @mapa has joined the channel
----
2019-07-21 17:58:28 UTC - mapa: does pulsar have altrnative for kstream
----
2019-07-21 17:58:29 UTC - mapa: ?
----
2019-07-21 18:00:59 UTC - David Kjerrumgaard: @mapa I would recommend looking 
at Heron Streamlets for this.  <https://streaml.io/blog/heron-streamlet-api>
----
2019-07-21 18:02:01 UTC - mapa: thank you
----
2019-07-21 18:02:19 UTC - mapa: @David Kjerrumgaard
----
2019-07-21 18:03:59 UTC - David Kjerrumgaard: @mapa No problem....It also 
supports joins.  
<https://apache.github.io/incubator-heron/docs/concepts/streamlet-api/>
----
2019-07-21 18:13:49 UTC - Jerry Peng: @mapa you can use Pulsar functions 
----
2019-07-21 18:14:58 UTC - Jerry Peng: @mapa 
<http://pulsar.apache.org/docs/en/functions-overview/>
----
2019-07-21 19:14:59 UTC - Jon Bock: @mapa Are you asking because you already 
have a KStreams application and you’re wondering how to port it to Pulsar, or 
is there a specific capability in KStreams that you’re looking for?  As Jerry 
says, most processing tasks can be handled by Pulsar Functions, or as David 
says there are a number of systems for stream processing that can be used with 
Pulsar if you have one of the scenarios that Pulsar functions can’t handle.
----
2019-07-22 01:56:51 UTC - mapa: not really im looking for a higher abstraction 
function that acts like kstream and ktable for ETL
----
2019-07-22 01:57:30 UTC - mapa: @Jon Bock we are in the process of testing out 
pulsar running side by side with our kafka
----
2019-07-22 01:58:23 UTC - mapa: since we are a GO and JAVA team we looked at 
what are the functions needed for use to migrate existing consumers to pulsar
----
2019-07-22 02:27:43 UTC - Jerry Peng: @mapa awesome! Pulsar Functions is great 
for etl workloads
----
2019-07-22 03:49:15 UTC - yansuopeng: @yansuopeng has joined the channel
----
2019-07-22 05:45:39 UTC - divyasree: Hi Sijieg, I am able to connect via proxy 
after giving different port number as suggested. Thanks :slightly_smiling_face:
----
2019-07-22 06:46:12 UTC - pradeep: Hi Team and @Matteo Merli,
How does the pulsar cluster (single logical cluster spread across multiple 
datacenter) behave in case of network breakage in between data-centers? Will 
there be split brain in pulsar cluster or pulsar handles this in different way?
----
2019-07-22 08:11:48 UTC - Sijie Guo: cool
----
2019-07-22 08:34:52 UTC - Sijie Guo: @pradeep if you add correct rack 
informations to bookies (you can use `bin/pulsar-admin bookies`), pulsar will 
place the data using rack-aware policy. It will then handle network breakage 
properly. If you have more than (include) 3 data centers, if one data center 
goes down, the pulsar cluster can still work as expect.
----
2019-07-22 08:41:59 UTC - pradeep: In case of 2 data centers , and replication 
factor 4 (rack aware, 2 replica in each DC), if one data center is down during 
write, will the write go through?
if yes, once the DC comes up, will it start replicating to maintain the 
replication factor.
----
2019-07-22 08:47:52 UTC - Sijie Guo: &gt; if one data center is down during 
write, will the write go through?

it is a complicated situation.

- the write can go through, since it will change ensemble.
- but if your broker crashes, it has to recover ledgers. recover require 
ack-quorum-size of bookies are alive. so if you set ack-quorum-size to be more 
than 2 (says 3), when a dc goes down, you can recover; but if you set 
ack-quorum-size to 2, it can lead to split-brain situation, which can cause 
data-loss during data recovery.
----

Reply via email to