Slack digest for #general - 2018-10-11

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

2018-10-10 13:03:05 UTC - John Davenport: @John Davenport has joined the channel
----
2018-10-10 13:03:50 UTC - John Davenport: Hey there!
----
2018-10-10 13:04:22 UTC - John Davenport: Id like to put pulsar on a project I'm working on
----
2018-10-10 13:04:51 UTC - John Davenport: I'm running all python asyncio
----
2018-10-10 13:05:50 UTC - John Davenport: Anyone know of a pulsar driver that supports asyncio
----
2018-10-10 15:43:29 UTC - Sijie Guo: @Penghui Li might be related to a recent change around handling native library in shading. @Matteo Merli thoughts?
----
2018-10-10 17:26:20 UTC - Matteo Merli: @John Davenport the Pulsar Python client library (<http://pulsar.apache.org/docs/en/client-libraries-python/>) is based off the C++ library.

In an async context, you should use `producer.send_async()` (<http://pulsar.apache.org/api/python/#pulsar.Producer.send_async>). That should work nicely with asyncio.
----
2018-10-10 17:27:46 UTC - Matteo Merli: I haven’t seen that error. The dependency on circe checksum has been removed from the shading profile, so it should show up as a Maven dependency
----
2018-10-10 21:20:23 UTC - Matthijs Van Der Welle: @Matthijs Van Der Welle has joined the channel
----
2018-10-10 21:45:46 UTC - Matt Jones: @Matt Jones has joined the channel
----
2018-10-10 23:00:34 UTC - Aaron Langford: @Aaron Langford has joined the channel
----
2018-10-10 23:09:56 UTC - Aaron Langford: Hey there. I'm an engineer from Instructure and I'm evaluating pulsar some of our data pipeline application needs.

Just wanting to see if there's a quick answer to my issue here (since I can't seem to find anything on Google). I'm playing with tiered storage, and I'm having some issues. So here are my questions:
1. I should expect auto offloading to kick off when the current segment is full. Does a segment correspond to a bookkeeper ledger?
2. When I kick off a manual offload, then check the status of the offload, I get the following:
```
Error in offload
null

Reason: Error offloading: org.apache.bookkeeper.mledger.ManagedLedgerException: java.util.concurrent.CompletionException: java.lang.UnsupportedOperationException
```
There seems to be nothing useful in any of the logs (`/opt/pulsar/logs`) as far as debugging this. Where should I go to figure out really what went wrong?

I am using a cluster on aws that was launched using the terraform provided in the incubator repo.
----
2018-10-10 23:10:06 UTC - Aaron Langford: Any help is greatly appreciated!
----
2018-10-10 23:13:30 UTC - Sijie Guo: @Aaron Langford :

&gt;  I should expect auto offloading to kick off when the current segment is full. Does a segment correspond to a bookkeeper ledger?

yes. the auto offloading is happening when rolling over ledgers.
a segment is a bookkeeper ledger.

&gt; java.lang.UnsupportedOperationException

exceptions seems to be indicating that offloaders are not installed and configured
----
2018-10-10 23:14:06 UTC - Sijie Guo: how do you deploy pulsar cluster?
----
2018-10-10 23:14:10 UTC - Aaron Langford: Cool, where are docs for installing s3 offloaders? If no docs, where is code?
----
2018-10-10 23:14:31 UTC - Sijie Guo: one second
----
2018-10-10 23:14:41 UTC - Aaron Langford: Cluster was deployed using terraform provided in the incubator repo.
----
2018-10-10 23:15:03 UTC - Sijie Guo: i see
----
2018-10-10 23:15:45 UTC - Sijie Guo: actually which version are you using?
----
2018-10-10 23:15:55 UTC - Sijie Guo: latest master?
----
2018-10-10 23:16:46 UTC - Aaron Langford: Ya, master as of Matteo Merli's commit on 2018-10-04 05:04:55
----
2018-10-10 23:19:37 UTC - Aaron Langford: And I am using the incubator repo as well (are you guys back to `apache/pulsar` with the recent promotion?)
----
2018-10-10 23:20:06 UTC - Matteo Merli: don’t worry about that. it’s the same repo, github redirects automatically
cool : Aaron Langford
----
2018-10-10 23:20:16 UTC - Sijie Guo: okay if you are using latest master’s terraform script, it will deploy 2.1.0-incubating binary.
then you can follow the following instructions to enable tiered storage: <http://pulsar.apache.org/docs/en/2.1.0-incubating/cookbooks-tiered-storage/#amazon-s3>
----
2018-10-10 23:25:52 UTC - Aaron Langford: So I'm happy to rebase onto latest master, but it doesn't appear that anything has changed around terraform ansible deploy stuff or the tiered storage components in the last 6 days (which is how old my local version is). So I'm a bit skeptical that would fix the issue.
----
2018-10-10 23:28:51 UTC - Sijie Guo: @Aaron Langford: the terraform scripts don’t do any tiered storage related deployment. when you are using tiered storage, you might need to modify the terraform scripts to do so. the tricky part is how to propagate your aws keys for accessing s3.

the doc here <http://pulsar.apache.org/docs/en/2.1.0-incubating/cookbooks-tiered-storage/#amazon-s3> describe the settings you need to change.
----
2018-10-10 23:29:55 UTC - Sijie Guo: we didn’t put that in terraform scripts, because it is very tricky to propagate aws credentials
----
2018-10-10 23:30:11 UTC - Sijie Guo: and we don’t want to provide a bad practice for people to follow
----
2018-10-10 23:33:08 UTC - Aaron Langford: Yes, so I have changed my broker.conf, and I have put in my own instance profile stuff for the ec2 instances. I can share that code if it helps.

Perhaps I did something incorrectly there, but I would expect credential issues to result in some exception being logged to one of my bookie machines. I'm going to ssh into my bookie and see if I can't hit that bucket via aws cli.
----
2018-10-10 23:39:40 UTC - Sijie Guo: @Aaron Langford yeah if you can share me with your configuration (hiding your keys), that would help me to understand the problem.
----
2018-10-10 23:41:44 UTC - Aaron Langford: iam_instance_profile is the only thing diff here. Granted to zookeeper, bookies, brokers. Not the proxy.
----
2018-10-10 23:42:51 UTC - Aaron Langford: defining the bucket for offloading
----
2018-10-10 23:43:01 UTC - Sijie Guo: Oh you didn’t update the deploy_pulsar.yaml?
----
2018-10-10 23:43:06 UTC - Aaron Langford: nah
----
2018-10-10 23:43:56 UTC - Aaron Langford: I did update templates/broker.conf
----
2018-10-10 23:44:09 UTC - Aaron Langford: That's the only non terraform thing I changed
----
2018-10-10 23:45:15 UTC - Aaron Langford: 
----
2018-10-10 23:45:19 UTC - Sijie Guo: Oh okay, can you share your template/broker.conf (hiding/removing your keys if you put something there)?
----
2018-10-10 23:47:27 UTC - Aaron Langford: The only thing changed in `broker.conf` is the three properties: `managedLedgerOffloadDriver` `s3ManagedLedgerOffloadRegion` and `s3ManagedLedgerOffloadBucket`
----
2018-10-10 23:49:58 UTC - Sijie Guo: I think I might know the reason, can you try change the driver from ‘aws-s3’ to S3?
----
2018-10-10 23:50:11 UTC - Aaron Langford: Would be happy to dig in more and maybe turn this into an issue on the repo if necessary. Just don't feel like I can even get enough information to put a ticket together.
----
2018-10-10 23:50:18 UTC - Aaron Langford: Yes, I can try that.
----
2018-10-10 23:50:55 UTC - Aaron Langford: :man-facepalming:
----
2018-10-10 23:51:01 UTC - Sijie Guo: Yes please file an issue for that. I can take a closer look after I get back to laptop 
----
2018-10-10 23:51:20 UTC - Aaron Langford: Not sure why I took down aws-s3 as the proper value there...
----
2018-10-10 23:51:25 UTC - Aaron Langford: Great catch.
----
2018-10-10 23:51:56 UTC - Aaron Langford: Thanks a bunch!
----
2018-10-10 23:52:09 UTC - Aaron Langford: That should fix things.
----
2018-10-10 23:52:23 UTC - Sijie Guo: There is a problem regarding documentation versioning 
----
2018-10-10 23:52:27 UTC - Sijie Guo: I guess 
----
2018-10-10 23:52:31 UTC - Aaron Langford: Would like to be involved in the future should evaluation continue to go well.
----
2018-10-10 23:52:46 UTC - Sijie Guo: If you run into any issues ping me
----
2018-10-10 23:52:57 UTC - Aaron Langford: So I'll keep my eye out for contribution opportunities.
----
2018-10-10 23:53:00 UTC - Sijie Guo: I will take a look when I get back to laptop 
----
2018-10-10 23:53:20 UTC - Aaron Langford: <https://pulsar.apache.org/docs/en/cookbooks-tiered-storage/>
----
2018-10-10 23:53:56 UTC - Aaron Langford: You are correct, that value is listed as `aws-s3` in this page, but as `S3` in the 2.1.0 docs you pointed me to.
----
2018-10-10 23:54:57 UTC - Matteo Merli: @Aaron Langford there was a problem with the 2.1.1 release that caused the docs from `master` to be published instead of the docs from the 2.1 branch
----
2018-10-10 23:55:33 UTC - Aaron Langford: Ah I see. Good to know.
----
2018-10-10 23:55:49 UTC - Aaron Langford: I'm signing off for the evening, but again, thanks for the help
+1 : Matteo Merli
----
2018-10-11 00:51:42 UTC - John Davenport: Anyone got experience implementing pulsar with python using asyncio?
----
2018-10-11 00:52:01 UTC - John Davenport: I've got a project I'd love to put it on.
----
2018-10-11 03:48:42 UTC - Sanjeev Kulkarni: @John Davenport can you elaborate what you mean by implementing pulsar with asyncio?
----
2018-10-11 03:48:49 UTC - Sanjeev Kulkarni: implement what?
----
2018-10-11 06:44:20 UTC - Sijie Guo: @John Davenport: pulsar python client is based on c++ client, it provides both sync and async api. if you use async api, it should work with asyncio.

examples: <http://pulsar.apache.org/api/python/#pulsar.Producer.send_async>
----