2019-03-28 09:50:42 UTC - Kirill Merkushev: @Kirill Merkushev has joined the channel ---- 2019-03-28 10:32:03 UTC - David Tinker: Is it safe to move the Bookkeeper ledgers directory to a different location in a live cluster? I will shutdown pulsar and bookkeeper on the node first, copy ledgers to its new location, update ledgerDirectories in bookkeeper.conf and then start them up again. I am worried that full paths to stuff in ledgers might appear in Zookeeper or the journal or something. thinking_face : David Tinker ---- 2019-03-28 10:53:43 UTC - Shivji Kumar Jha: It seems there is a problem with --output option in ```bin/pulsar tokens create-secret-key --output my-secret.key```
This generates some encoded bytes on stdout: ```bin/pulsar tokens create-secret-key``` But --ouput does not generate a file at all. ---- 2019-03-28 10:55:32 UTC - Shivji Kumar Jha: Doc reference: <https://pulsar.apache.org/docs/en/security-token-admin/#creating-a-secret-key> ---- 2019-03-28 10:57:00 UTC - Alexandre DUVAL: It's generating it at the root dir of pulsar. ---- 2019-03-28 10:57:08 UTC - Shivji Kumar Jha: let me see ---- 2019-03-28 10:57:29 UTC - Alexandre DUVAL: You can define an absolute path. ---- 2019-03-28 10:58:03 UTC - Shivji Kumar Jha: It is indeed there at the root directory of pulsar - thanks @Alexandre DUVAL ---- 2019-03-28 11:50:45 UTC - Shubham Maheshwari: @Shubham Maheshwari has joined the channel ---- 2019-03-28 11:55:03 UTC - Shubham Maheshwari: Hey Guys .. I am using the pulsar java client lib (v2.3.0) I am creating multiple consumers for a subscription in shared mode, but the messages are getting delivered to only one (and not to the rotating to the next idle consumer) following is the code to create the consumers: ---- 2019-03-28 11:55:25 UTC - Shubham Maheshwari: ---- 2019-03-28 11:55:34 UTC - Shubham Maheshwari: can anyone help please. ---- 2019-03-28 12:03:35 UTC - Shubham Maheshwari: I tried the same with lib v2.2.0 .. its working fine ---- 2019-03-28 12:37:48 UTC - Darragh: so with more experiments to figure out where tail latency spikes are coming from I've noticed that writing the journal to 2 locations (on the same disk) instead of 1 seems to half the spikes. This seems to tell me that the problem lies with the journal flushes? Can I tune some parameter to do this more frequently maybe, so that the spikes are less outspoken (at the possible cost of a general higher normal latency) or is there another way to soften the spikes ? ---- 2019-03-28 12:38:52 UTC - Steve Kim: <https://github.com/apache/pulsar/pull/3929> +1 : jia zhai clap : jia zhai ---- 2019-03-28 14:36:26 UTC - Ezequiel Lovelle: Hi @Shubham Maheshwari please try to put `+ i` variable part on `subscriptionName` instead of `consumerName` ---- 2019-03-28 14:37:44 UTC - Ezequiel Lovelle: please take a look at the comment above. ---- 2019-03-28 15:51:45 UTC - Ryan Samo: Hey guys, I have a scenario that I need some quick advice on if someone could lend their expertise? If you have... - 1 partitioned topic - Message retention = 24 hours - Message expiry = 24 hours - 1 active consumer subscription Can you... - Create another subscription to read messages from N to N offset? Like if you wanted to basically replay some messages that had already been acknowledged in the past 24 hours but without interrupting the consumer that is already in place? - I see from the admin side of Pulsar, the API has "Reset subscription to message position closest to absolute timestamp (in ms)." Can we use that on the 2nd subscription to place the other consumer close enough to replay the already acknowledged messages? The reader interface would be ideal but we have a partitioned topic in this case. Just trying to figure out if this is even possible? Thanks! ---- 2019-03-28 15:54:04 UTC - Matteo Merli: You can create a reader for each partition ---- 2019-03-28 15:55:09 UTC - Matteo Merli: Use `client.getPartitionsForTopic()` to get the list of partitions ---- 2019-03-28 15:55:35 UTC - Ryan Samo: Oh interesting! ---- 2019-03-28 15:56:21 UTC - Ryan Samo: Is there a good reader example on the github or elsewhere that would be a good starting point? ---- 2019-03-28 15:56:46 UTC - Matteo Merli: Not sure :) ---- 2019-03-28 15:57:16 UTC - Matteo Merli: But you can treat each partition as a “non-partitioned” topic ---- 2019-03-28 15:57:53 UTC - Ryan Samo: Ok perfect, sounds like a good science experiment on my end. Thanks @Matteo Merli ---- 2019-03-28 16:02:29 UTC - Emma Pollum: Morning guys! Last night, our zookeeper disks ran full, and after shutting down and growing the disks we had full functionality, but none of our messages were in the system. Any ideas on how to recover them, or how to prevent this in the future? ---- 2019-03-28 16:08:45 UTC - Matteo Merli: Is the ZK data recovered? ---- 2019-03-28 16:09:24 UTC - Emma Pollum: It appears to be. My pulsar setup is producing and consuming new messages ---- 2019-03-28 16:10:20 UTC - Emma Pollum: Where is topic data stored? The namespaces and tenant information was all solid, but no topics showed up until I started producing to them again ---- 2019-03-28 16:11:02 UTC - Matteo Merli: That would be under `/managed-ledgers/...` ---- 2019-03-28 16:11:54 UTC - Emma Pollum: I definitely see that my bookies data still has a significant number of entries, so my messages are still there, just not in a way zk can see? ---- 2019-03-28 16:14:00 UTC - Matteo Merli: But is the topic metadata completely missing? Did ZK come back up with a consistent state or it was rolled back to an earlier snapshot ? ---- 2019-03-28 16:15:05 UTC - Emma Pollum: All we did was shut down the service, grow the disks and remount them. No rollback. ---- 2019-03-28 16:16:14 UTC - Ryan Samo: Sorry @Matteo Merli, I’m sure you’re questioned out by now lol... and please help @Emma Pollum 1st. If you have a cluster of bookies, say 6 of them, and you have a write quorum of 2, what happens if the underlying storage fills up? Also, if you are calculating available storage to Bookkeeper, would you sum all of the drives together , 6 servers with 100G each = 600G available or do we use 1 server x 100G = 100G Max? ---- 2019-03-28 16:16:16 UTC - Matteo Merli: Uhm, then I don’t know why they would disappear after restart ---- 2019-03-28 16:16:33 UTC - Emma Pollum: :disappointed: me either ---- 2019-03-28 16:23:47 UTC - Chris DiGiovanni: @Emma Pollum did you try bouncing your broker(s)? ---- 2019-03-28 16:25:02 UTC - Chris DiGiovanni: I had something similar happen to me when trying to deploy k8s. Though it was early on in my testing and can't remember if that resolved it or not. ---- 2019-03-28 16:29:51 UTC - Dave Southwell: Colleague of @Emma Pollum's here. Yes @Chris DiGiovanni we did. ---- 2019-03-28 16:31:20 UTC - Chris DiGiovanni: :confused: ---- 2019-03-28 16:32:01 UTC - Chris DiGiovanni: I guess the next thing I would start doing is digging into the bookkeeper shell cmd to see if it thinks the data is there or not. ---- 2019-03-28 16:42:55 UTC - Emma Pollum: I was able to see that it seems our messages are there, but the ledgers are marked "CLOSED" ---- 2019-03-28 16:42:56 UTC - Emma Pollum: ledgerID: 00000000-0000-0000-0000-00000000c023 BookieMetadataFormatVersion 2 quorumSize: 2 ensembleSize: 2 length: 3971383 lastEntryId: 49999 state: CLOSED segment { ensembleMember: "10.128.0.10:3181" ensembleMember: "10.128.0.185:3181" firstEntryId: 0 } digestType: CRC32C password: "" ackQuorumSize: 2 ---- 2019-03-28 16:43:22 UTC - Emma Pollum: Can you help me understand what happens when a ledger is CLOSED? ---- 2019-03-28 16:44:42 UTC - Chris DiGiovanni: ``` After the ledger has been closed---either explicitly or because the writer process crashed---it can then be opened only in read-only mode. ``` ---- 2019-03-28 16:46:47 UTC - Chris DiGiovanni: I just started digging into Pulsar around 3 Weeks ago. So this is fairly new to me as well. Does the bookkeeper cluster look in a healthy state. No odd logs, bookeeper shell returns both as available? ---- 2019-03-28 17:03:34 UTC - Emma Pollum: Yeah, now that I dig into more of the bookkeeper ledger meta, i have at least one ledger that has a different ensemble size and quorum size. Also a lot of the ledgers have the same "last entry id" ---- 2019-03-28 17:20:47 UTC - Emma Pollum: Looks like one of our global zks is out with the error "Mar 28 17:19:21 p-zookeeper-3 zkServer.sh[18195]: Caused by: java.io.IOException: Unreasonable length = 1527340800" ---- 2019-03-28 17:43:11 UTC - bhagesharora: Hello everyone, I am trying to implement pulsar-kafka-adaptor program through java Reference URL - <https://github.com/apache/pulsar/blob/master/pulsar-client-kafka-compat/pulsar-client-kafka-tests/pom.xml> I had given maven build But its giving error, see below [ERROR] Failed to execute goal on project pulsar_kafka_adaptor: Could not resolve dependencies for project pulsar_kafka_adaptor:pulsar_kafka_adaptor:jar:0.0.2-SNAPSHOT: Failed to collect dependencies for pulsar_kafka_adaptor:pulsar_kafka_adaptor:jar:0.0.2-SNAPSHOT: Could not resolve version conflict among [org.apache.pulsar:pulsar-broker:jar:2.3.0 -> org.apache.pulsar:pulsar-zookeeper-utils:jar:2.3.0 -> org.apache.bookkeeper:stream-storage-server:jar:4.9.0 -> org.apache.bookkeeper:stream-storage-java-client:jar:4.9.0 -> io.grpc:grpc-core:jar:1.12.0, org.apache.pulsar:pulsar-broker:jar:2.3.0 -> org.apache.pulsar:pulsar-zookeeper-utils:jar:2.3.0 -> org.apache.bookkeeper:stream-storage-server:jar:4.9.0 -> org.apache.bookkeeper:stream-storage-java-client:jar:4.9.0 -> io.grpc:grpc-protobuf-lite:jar:1.12.0 -> io.grpc:grpc-core:jar:1.12.0, org.apache.pulsar:pulsar-broker:jar:2.3.0 -> org.apache.pulsar:pulsar-zookeeper-utils:jar:2.3.0 -> org.apache.bookkeeper:stream-storage-server:jar:4.9.0 -> org.apache.bookkeeper:stream-storage-java-client:jar:4.9.0 -> io.grpc:grpc-testing:jar:1.12.0 -> io.grpc:grpc-core:jar:[1.12.0,1.12.0], org.apache.pulsar:pulsar-broker:jar:2.3.0 -> org.apache.pulsar:pulsar-zookeeper-utils:jar:2.3.0 -> io.grpc:grpc-all:jar:1.18.0 -> io.grpc:grpc-auth:jar:1.18.0 -> io.grpc:grpc-core:jar:[1.18.0,1.18.0], org.apache.pulsar:pulsar-broker:jar:2.3.0 -> org.apache.pulsar:pulsar-zookeeper-utils:jar:2.3.0 -> io.grpc:grpc-all:jar:1.18.0 -> io.grpc:grpc-core:jar:[1.18.0,1.18.0], org.apache.pulsar:pulsar-broker:jar:2.3.0 -> org.apache.pulsar:pulsar-zookeeper-utils:jar:2.3.0 -> io.grpc:grpc-all:jar:1.18.0 -> io.grpc:grpc-netty:jar:1.18.0 -> io.grpc:grpc-core:jar:[1.18.0,1.18.0], org.apache.pulsar:pulsar-broker:jar:2.3.0 -> org.apache.pulsar:pulsar-zookeeper-utils:jar:2.3.0 -> io.grpc:grpc-all:jar:1.18.0 -> io.grpc:grpc-okhttp:jar:1.18.0 -> io.grpc:grpc-core:jar:[1.18.0,1.18.0], org.apache.pulsar:pulsar-broker:jar:2.3.0 -> org.apache.pulsar:pulsar-zookeeper-utils:jar:2.3.0 -> io.grpc:grpc-all:jar:1.18.0 -> io.grpc:grpc-protobuf:jar:1.18.0 -> io.grpc:grpc-core:jar:1.18.0, org.apache.pulsar:pulsar-broker:jar:2.3.0 -> org.apache.pulsar:pulsar-zookeeper-utils:jar:2.3.0 -> io.grpc:grpc-all:jar:1.18.0 -> io.grpc:grpc-protobuf-nano:jar:1.18.0 -> io.grpc:grpc-core:jar:1.18.0, org.apache.pulsar:pulsar-broker:jar:2.3.0 -> org.apache.pulsar:pulsar-zookeeper-utils:jar:2.3.0 -> io.grpc:grpc-all:jar:1.18.0 -> io.grpc:grpc-stub:jar:1.18.0 -> io.grpc:grpc-core:jar:1.18.0] -> [Help 1] what would be the solution ?? ---- 2019-03-28 18:06:58 UTC - Vikas: hey @David Kjerrumgaard, how are you! I am unable to send the data to the pulsar topic using NiFi and getting the error as : Policies not found for alpha/xxx-xxxxx namespace Do you know what am I missing? ---- 2019-03-28 18:10:26 UTC - David Kjerrumgaard: Hi @Vikas I am OOO this week but can take a look this weekend. Can you share the pulsar and nifi logs? ---- 2019-03-28 18:11:15 UTC - David Kjerrumgaard: Did you configure any polices on that namespace? ---- 2019-03-28 18:11:32 UTC - Vikas: sure, I can share with you the NiFi logs but dont have access to the pulsar logs as its the production system. I can definitely ask the person who takes care of the pulsar system ---- 2019-03-28 18:12:12 UTC - David Kjerrumgaard: The issue appears to be related to Pulsar policies ---- 2019-03-28 18:12:31 UTC - Vikas: oh ok, let me ask him. Sorry to bug you on your holidays. ---- 2019-03-28 18:12:37 UTC - Vikas: enjoy :smile: ---- 2019-03-28 20:44:17 UTC - Devin G. Bost: @Devin G. Bost has joined the channel ---- 2019-03-28 20:44:44 UTC - Grant Wu: Are the Python docs at <https://pulsar.apache.org/api/python> out of date? ---- 2019-03-28 20:45:15 UTC - Grant Wu: Poring over the diffs between the 2.2.1 tag and the 2.3.0 tag, I see this: ``` diff --git a/pulsar-client-cpp/python/src/message.cc b/pulsar-client-cpp/python/src/message.cc index 42322490e..19aa7bb2b 100644 --- a/pulsar-client-cpp/python/src/message.cc +++ b/pulsar-client-cpp/python/src/message.cc @@ -66,6 +66,20 @@ boost::python::object Message_data(const Message& msg) { return boost::python::object(boost::python::handle<>(PyBytes_FromStringAndSize((const char*)msg.getData(), msg.getLength()))); } +boost::python::object Message_properties(const Message& msg) { + boost::python::dict pyProperties; + for (const auto& item : msg.getProperties()) { + pyProperties[item.first] = item.second; + } + return pyProperties; +} + +std::string Topic_name_str(const Message& msg) { + std::stringstream ss; + ss << msg.getTopicName(); + return ss.str(); +} + const MessageId& Message_getMessageId(const Message& msg) { return msg.getMessageId(); } @@ -109,7 +123,7 @@ void export_message() { ; class_<Message>("Message") - .def("properties", &Message::getProperties, return_value_policy<copy_const_reference>()) + .def("properties", &Message_properties) .def("data", &Message_data) .def("length", &Message::getLength) .def("partition_key", &Message::getPartitionKey, return_value_policy<copy_const_reference>()) @@ -117,5 +131,6 @@ void export_message() { .def("event_timestamp", &Message::getEventTimestamp) .def("message_id", &Message_getMessageId, return_value_policy<copy_const_reference>()) .def("__str__", &Message_str) + .def("topic_name", &Topic_name_str) ; } ``` ---- 2019-03-28 20:45:19 UTC - Matteo Merli: They shouldn’t. Do you see anything missing thee? ---- 2019-03-28 20:45:29 UTC - Grant Wu: But the `topic_name` function isn’t in the docs ---- 2019-03-28 20:45:36 UTC - Grant Wu: Refer to thread ---- 2019-03-28 20:47:53 UTC - Devin G. Bost: My team is trying to figure out how to delete subscriptions. We're getting: ``` Failed to perform http delete request: javax.ws.rs.ClientErrorException: HTTP 412 Precondition Failed Subscription has active connected consumers Reason: Subscription has active connected consumers ``` We've tried terminating the topic first, but we're still missing something, and we can't find anything in the docs about it. ---- 2019-03-28 20:49:06 UTC - Matteo Merli: If a consumer is connected, the deletion fails. That’s Because the consumers will immediately try to reconnect and recreate the subscription ---- 2019-03-28 20:49:24 UTC - Devin G. Bost: Gotcha. How do we disconnect the consumers? ---- 2019-03-28 20:50:05 UTC - Devin G. Bost: Is there a forceful way to do that? We're trying to wipe a deployment and re-deploy via automation. ---- 2019-03-28 20:50:42 UTC - Matteo Merli: There a way to force the topic deletion ---- 2019-03-28 20:51:27 UTC - Matteo Merli: Disconnecting consumer is just shutting down the application is consuming :) ---- 2019-03-28 20:52:23 UTC - Devin G. Bost: There is, or there is not a way to force the topic deletion? Sorry, I'm not understanding your comment. ---- 2019-03-28 21:00:57 UTC - Devin G. Bost: It looks like our deletions are only breaking on: ``` <persistent://public/functions/coordinate> <persistent://public/functions/assignments> <persistent://public/functions/metadata> ``` So, maybe my question is moot. ---- 2019-03-28 21:01:46 UTC - Matteo Merli: These are topics use by pulsar functions runtime. +1 : Devin G. Bost ---- 2019-03-28 21:08:04 UTC - Grant Wu: @Matteo Merli try `git diff tags/v2.2.1 tags/v2.3.0 -- pulsar-client-cpp/python/pulsar` ---- 2019-03-28 21:08:42 UTC - Grant Wu: and that should get a bunch of stuff missing… ---- 2019-03-28 21:23:02 UTC - Grant Wu: Anyone know how to build the Python docs? ---- 2019-03-28 21:27:48 UTC - Ali Ahmed: @bhagesharora seems like “io.grpc:grpc-core:jar:1.12.0” “io.grpc:grpc-all:jar:1.18.0" is the issue ---- 2019-03-28 21:28:08 UTC - Ali Ahmed: is this in master ? or some new modified pom ---- 2019-03-28 21:52:16 UTC - Florian Hof: @Florian Hof has joined the channel ---- 2019-03-28 21:58:04 UTC - Florian Hof: How big can a message be? Is 500MB ok? Is Pulsar suitable as a document-based database with long retention (more than one year)? Thanks for feedback :slightly_smiling_face: ---- 2019-03-28 21:58:39 UTC - Ali Ahmed: @Florian Hof messages size is currently limited to 5MB ---- 2019-03-28 22:01:18 UTC - Ali Ahmed: @bhagesharora this may help ``` <!--Adding dependencyManagement to resolve conflicts --> <dependencyManagement> <dependencies> <dependency> <groupId>io.grpc</groupId> <artifactId>grpc-core</artifactId> <version>1.2.0</version> </dependency> </dependencies> </dependencyManagement> ``` ---- 2019-03-28 22:02:19 UTC - Grant Wu: I’m not sure why you would want to use Pulsar as a document-based database ---- 2019-03-28 22:07:43 UTC - Emma Pollum: After a crashed zk, we are left without access to messages that are still in bookkeeper. Does anyone have any thoughts on the best way to extract them? Essentially, all the data in the bookeeper is fine, but there is no reference to it under /managed-ledgers in zk. ---- 2019-03-28 22:09:16 UTC - Florian Hof: @Grant Wu I was looking for a messaging system with long-term storage for big messages (100MB). Ok, Pulsar is definitly not done for that. ---- 2019-03-28 22:09:54 UTC - Grant Wu: It’s not clear to me why you wouldn’t use a traditional database for this use case ---- 2019-03-28 22:11:54 UTC - Ryan Samo: Do Pulsar clients cache DNS? Meaning if you have DNS being used between all of the server nodes, brokers, etc., will the clients cache the IP of the last broker they talked to? Just thinking in terms of if you have an outage on a broker and the client cached the IP, it can take up the TTL of the DNS before a client figures that out. Thoughts? ---- 2019-03-28 22:14:26 UTC - Grant Wu: Hey, @Laurent Chriqui are you around? ---- 2019-03-28 22:14:36 UTC - Grant Wu: I’m running into ``` ImportError: dlopen(/Users/grant.wu/petuum/sps_venv/lib/python3.7/site-packages/_pulsar.cpython-37m-darwin.so, 2): Symbol not found: __ZN6icu_636LocaleC1ERKS0_ Referenced from: /Users/grant.wu/petuum/sps_venv/lib/python3.7/site-packages/_pulsar.cpython-37m-darwin.so Expected in: flat namespace in /Users/grant.wu/petuum/sps_venv/lib/python3.7/site-packages/_pulsar.cpython-37m-darwin.so ``` ---- 2019-03-28 22:14:41 UTC - Grant Wu: and I think you encountered this before? ---- 2019-03-28 22:15:35 UTC - Grant Wu: I think it’s in this thread <https://lists.apache.org/thread.html/63f3ce8255958d30d3ee75c2e63bc04ced1da65d7e4c1e522fc29b27@%3Cusers.pulsar.apache.org%3E> ---- 2019-03-28 22:17:13 UTC - Grant Wu: I’m going to try `brew upgrade` ---- 2019-03-28 22:17:16 UTC - Grant Wu: see if that fixes anything ---- 2019-03-28 22:32:13 UTC - Grant Wu: ah okay I need the `.post1` version ok_hand : Laurent Chriqui ---- 2019-03-28 22:43:49 UTC - Matteo Merli: Uhm, you’re right. that’s way off :? ---- 2019-03-28 22:43:52 UTC - Matteo Merli: :confused: ---- 2019-03-28 22:45:17 UTC - Matteo Merli: <https://github.com/apache/pulsar/tree/asf-site/content/api/python> ---- 2019-03-28 22:45:43 UTC - Matteo Merli: It looks like the website build is not updating the docs correctly ---- 2019-03-28 22:53:49 UTC - Grant Wu: Is there an easy way to build them locally? ---- 2019-03-28 22:54:15 UTC - Matteo Merli: There’s a command that builds it on a docker image ---- 2019-03-28 22:54:22 UTC - Matteo Merli: let me search for it ---- 2019-03-28 22:54:56 UTC - Matteo Merli: essentially it’s calling the `pdoc` tool ---- 2019-03-28 22:55:09 UTC - Matteo Merli: `site2/tools/docker-build-site.sh` ---- 2019-03-28 22:55:41 UTC - Matteo Merli: `site2/tools/publish-website.sh` it’s the one that does the publish.. which is probably where it’s currently broken ---- 2019-03-28 23:23:21 UTC - Grant Wu: I tried to build it in docker and I got ``` c++: internal compiler error: Killed (program cc1plus) ``` ---- 2019-03-28 23:23:39 UTC - Grant Wu: I imagine this is probably an OOM? ---- 2019-03-28 23:23:57 UTC - Grant Wu: Surely it should be possible to build the docs without building the world ;-; ---- 2019-03-28 23:26:34 UTC - Ezequiel Lovelle: afaik, java client uses connection pool to handle connections towards every broker-topic. Dns resolution will occur (the first time) when calling `create()` for producers and `subscribe()` for consumers and no cache is involved on dns resolution I think. So, clients will try to do re-connection on broker outage, and this process will start again doing new dns resolution. But please, don't take this to be final and wait if anyone have more information about this. ---- 2019-03-28 23:28:53 UTC - Ryan Samo: Ok thanks @Ezequiel Lovelle , this is the behavior I see in my testing so looking for a sanity check from the community. I’ll wait and see if anyone else adds information +1 : Ezequiel Lovelle ---- 2019-03-28 23:41:11 UTC - Grant Wu: @Matteo Merli Is there… any sort of documentation as to how to use JSON Schemas with the pulsar 2.3.0 client? I’m looking at `definition.py` and it seems like it’s using a metaclass approach? Personally all I want is to ser/deser stuff to dictionaries.. at least for now ---- 2019-03-28 23:41:25 UTC - Matteo Merli: :slightly_smiling_face: ---- 2019-03-28 23:41:55 UTC - Matteo Merli: it’s building the C++ library to have the _pulsar.so to be imported ---- 2019-03-28 23:42:10 UTC - Grant Wu: <https://pulsar.apache.org/docs/en/concepts-schema-registry/> says “The Pulsar schema registry is currently available only for the Java client.” - is this still the case? ---- 2019-03-28 23:43:04 UTC - Matteo Merli: @Grant Wu <https://pulsar.apache.org/docs/en/client-libraries-python/#schema> ---- 2019-03-28 23:43:32 UTC - Matteo Merli: yes.. the above line is now inaccurate ---- 2019-03-28 23:44:25 UTC - Grant Wu: Oh, right, I forgot there were non-generated Python docs… ---- 2019-03-28 23:48:23 UTC - Grant Wu: If I have a client that doesn’t know anything about schemas - say, it uses the websocket interface, or the Go client - will problems arise if it tries to consume from a topic with a schema? ---- 2019-03-28 23:49:42 UTC - Grant Wu: If I want to use a purely client-side approach for now, can I just pass in `None` for the third parameter to the Schema class constructor? ---- 2019-03-28 23:49:55 UTC - Grant Wu: as `BytesSchema` and `StringSchema` are doing ---- 2019-03-29 01:47:59 UTC - jia zhai: Thanks @Steve Kim for the work. ---- 2019-03-29 04:40:33 UTC - Yuvaraj Loganathan: @Emma Pollum What do you mean by crashed zk ? ZK persists all the data to disk. rebooting zk should solve or did not bring the data back? how many nodes are there in zk cluster ? ---- 2019-03-29 05:35:50 UTC - David Tinker: Can the Bookkeeper ledgers directory configured in bookkeeper.conf be a symlink to a directory on a different filesystem? ---- 2019-03-29 05:39:43 UTC - Sijie Guo: @David Tinker yes +1 : David Tinker, Guangzhong Yao ---- 2019-03-29 06:07:25 UTC - Shubham Maheshwari: Hi @Ezequiel Lovelle.. if I do that, I'll end up creating multiple subscriptions to the topic. I do not want that. I want a single subscription with multiple consumer threads running for it ---- 2019-03-29 06:07:47 UTC - Shubham Maheshwari: btw .. this works fine in the previous version. ----
