Slack digest for #general - 2020-04-05

[Apache Pulsar Slack <ap...@gmail.com>]

2020-04-04 15:40:26 UTC - steven meadows: @steven meadows has joined the channel
----
2020-04-04 15:42:11 UTC - steven meadows: I'm interested in integrating Apache Pulsar schema registry, but have question. Can Pulsar Schema Registry support multi-tenacy?
----
2020-04-04 18:12:54 UTC - Shangpeng Sun: @Shangpeng Sun has joined the channel
----
2020-04-04 18:18:09 UTC - Shangpeng Sun: hello, can someone please help me understand how selective ack works in pulsar? as I understand the acks are stored in the cursor ledger, doesn't pulsar have to replay the entire history of the cursor ledger of this subscription to figure out where. the gaps are?
----
2020-04-05 02:55:04 UTC - Franck Schmidlin: @Franck Schmidlin has joined the channel
----
2020-04-05 05:01:01 UTC - Matteo Merli: Whenever we update the cursor ledger, we append the current state of the cursor, which includes the mark-delete position (for which everything before that was already deleted) and a rangeset if the selective acks. 

On recovery, we only need to read the last entry in the cursor ledger to restore the state.  
----
2020-04-05 05:01:48 UTC - Matteo Merli: Yes the permissions on the schema are attached to the topic permissions
----
2020-04-05 05:16:54 UTC - Shangpeng Sun: Cool, thanks that makes sense. For how volume subscription with selective acks tho, wouldn’t it cause the last entry too large, potentially contain thousands or even millions of records? If I have two consumer in a shared subscription and intentionally only make one consumer to ack, the size of the range set will be unbounded?
----
2020-04-05 07:45:35 UTC - Franck Schmidlin: Hi everyone. Here is a question I started on twitter before being redirected here.

I'm undergoing an interesting mental journey. Started last week after reading a free ebook on event driven micro services architectures and kafka.   
A *lot* of googling later, I am now seriously considering pulsar .
What bugs me is that I want to support a Serverless architecture, on @awscloud , and it feels counterintuitive to start with a bunch of servers to run the clusters.
There is managed service @apachekafka , but doesn't seem easier than managing the cluster myself.
What I really need to get my head around is whether I could run containerised  zookeepers, brokers and bookies as #aws #fargate tasks.

Not EKS, can't deal with that agro.

I don't have long term event persistence requirement just yet. And if/when I do I could run the bookies as @awscloud EC2 tasks backed by EBS, no?

My main requirement is elasticity. Run as small a resilient infrastructure I can, but scale it up to deal with spike in customer requests. 

My main inspiration is this blog <https://www.splunk.com/en_us/blog/it/comparing-pulsar-and-kafka-how-a-segment-based-architecture-delivers-better-performance-scalability-and-resilience.html|https://www.splunk.com/en_us/blog/it/comparing-pulsar-and-kafka-how-a-segment-based-architecture-delivers-better-performance-scalability-and-resilience.html>

but I need to translate it into a deployment diagram. 

Anybody reading this who can point me to any more info about any of this?
I'm a bit out of my depth :weary:

Thanks
----