2020-07-30 10:02:49 UTC - Adriaan de Haan: Hi, I have a pulsar 2.6.0 
installation and for some reason some acknowledged messages are remaining in 
the backlog (in the stat of the subscription from the admin interface),  what 
are the conditions that could result in that happening? It is a persistent 
topic with a single shared subscription with a few consumers, all of them 
ack'ing right after receiving the message.
----
2020-07-30 10:50:49 UTC - Ebere Abanonu: @Sijie Guo Pulsar need to fix Presto 
integration. It is unreliable or is there a way to make it better via 
configuration?
----
2020-07-30 11:10:51 UTC - Mohit Manaskant: Does `pulsar functions` suppport 
`key_shared` subscription mode?
----
2020-07-30 11:12:27 UTC - Mohit Manaskant: Looks like not as per the issue 
here: <https://github.com/apache/pulsar/issues/6527>
----
2020-07-30 12:25:58 UTC - Joshua Decosta: my understanding is that disabling 
deduplication allows only new messages to be sent to the topic. It could 
confuse the consumers to see the same message over and over again.
----
2020-07-30 12:29:37 UTC - Ebere Abanonu: I guess sql-worker should only run 
when brokers and bookies are up and running
----
2020-07-30 15:43:47 UTC - Aaron: How do you build a pulsar generic record to 
send in the java client? The example here at the bottom is incorrect 
(<https://pulsar.apache.org/docs/en/concepts-schema-registry/|https://pulsar.apache.org/docs/en/concepts-schema-registry/>)
 . schema.newRecordBuilder is invalid
----
2020-07-30 16:01:54 UTC - Maximilian Michels: @Maximilian Michels has joined 
the channel
----
2020-07-30 16:03:32 UTC - Ebere Abanonu: @Sijie Guo so pulsar topic is case 
sensitive? Students and students is not same topic?
----
2020-07-30 16:06:14 UTC - Sijie Guo: What is your behavior you observed?
----
2020-07-30 16:10:39 UTC - Sijie Guo: @Takahiro Hozumi the message deduplication 
requires additional setup at the producer side. It is also has additional 
overhead a) it uses a de-dup cursor for handling de-duplication b) it requires 
additional fail handling logic when a topic is loaded by broker. Hence we 
disable it by default.
----
2020-07-30 16:12:50 UTC - Chris Hansen: It seems unusual for topic stats to say 
3 messages consumed while function stats says 0 received. Is that expected that 
these can be out of sync in certain cases?
----
2020-07-30 16:13:14 UTC - Sijie Guo: You can run `topics stats-internal` to see 
if the individualDeleteMsgs is growing. If so, it typically means missing acks.

Then if you run `topics unload` to unload the topic. If the messages are 
redelivered and backlog is cleared, that usually means two things:

a) you might not have ackTimeout. Hence the messages are not redelivered if 
acks are missed.
b) there might be some issues in consumer group acking logic.
----
2020-07-30 16:13:28 UTC - Sijie Guo: It is not supported yet.
----
2020-07-30 16:15:25 UTC - Sijie Guo: You can check this code example: 
<https://github.com/streamnative/pulsar-examples/blob/master/clients/schema/src/main/java/io/streamnative/examples/schema/generic/GenericSchemaProducerExample.java>
----
2020-07-30 16:19:15 UTC - Sijie Guo: I don’t think so. unless function worker 
resets the stats. What is your function doing? Does it write results to 
somewhere else?
----
2020-07-30 16:21:56 UTC - Chris Hansen: results are meant to go to an output 
topic, but they never get there. It also logs to the context logger and 
System.out . Very simple.

```
public class SecretPrintingFunction implements Function&lt;String, String&gt; {
    private static final Logger LOGGER = 
LoggerFactory.getLogger(SecretPrintingFunction.class);

    public SecretPrintingFunction() {
        log(LOGGER, "new SecretPrintingFunction() created");
    }

    @Override
    public String process(String input, Context context) {
        log(context.getLogger(), "Input: " + input);
        final String secret = context.getSecret(input);
        log(context.getLogger(), "Secret: " + secret);
        return secret == null ? "null" : secret; //returning actual null 
indicates no output
    }

    private void log(Logger logger, String msg) {
        <http://LOGGER.info|LOGGER.info>(msg);
        <http://logger.info|logger.info>(msg);
        System.out.println(msg);
    }
}```
----
2020-07-30 16:23:07 UTC - Chris Hansen: function config:
```tenant: "public"
namespace: "default"
name: "SecretPrintingFunction"
jar: "/app/lib/pulsar-vault-secrets-test.jar"
className: "com.overstock.pulsar.SecretPrintingFunction"
inputs: ["<persistent://public/default/test-input>"]
output: "<persistent://public/default/test-output>"
secrets:
  one: "/secret-printing-function/one"
  two: "/secret-printing-function/two"
  three: "/secret-printing-function/three"```
----
2020-07-30 16:23:56 UTC - Chris Hansen: i’m working on a custom 
`SecretsProvider`  and the function is just for testing
----
2020-07-30 16:29:03 UTC - Ebere Abanonu: Yesterday I experienced the issue 
where Presto was throwing No table found exception when querying data in the 
table. However when I query for the tables in the schema it retrieves 
successfully. Priorly I set the retention to -1, -1 that is when I started 
experiencing the issue of no table found
----
2020-07-30 16:29:48 UTC - Ebere Abanonu: Then today after publishing to a 
topic, presto was returning empty set. I changed the order of deployment to 
happen after broker is fully up and running to get it working properly
----
2020-07-30 17:34:19 UTC - Kalyn Coose: Thanks. Perhaps I negated my ability to 
see message chunking working when I set my maxMessageSize to 20 MB 
:face_with_hand_over_mouth:
----
2020-07-30 17:38:28 UTC - Kalyn Coose: Why am I unable to find 
org.apache.pulsar.functions? I have Pulsar 2.6.0 dependency installed
----
2020-07-30 17:40:00 UTC - Chris Hansen: there are different dependencies, you 
need pulsar-functions-api
```    &lt;dependency&gt;
      &lt;groupId&gt;org.apache.pulsar&lt;/groupId&gt;
      &lt;artifactId&gt;pulsar-functions-api&lt;/artifactId&gt;
    &lt;/dependency&gt;```
----
2020-07-30 17:40:58 UTC - Kalyn Coose: Sorry I didn't see that, is that missing 
information on Pulsar docs?
----
2020-07-30 17:41:48 UTC - Chris Hansen: not that I saw
----
2020-07-30 17:43:28 UTC - Kalyn Coose: I've been reading the docs and I just 
didn't notice anything about a separate dependency. Thanks for your help. 
<https://pulsar.apache.org/docs/en/functions-overview/>
----
2020-07-30 17:44:12 UTC - Chris Hansen: sure thing :thumbsup:
----
2020-07-31 02:05:21 UTC - Adriaan de Haan: Does this tell you anything? I can't 
really interpret these values yet:
```
  "entriesAddedCounter" : 704,
  "numberOfEntries" : 26,
  "totalSize" : 21490,
  "currentLedgerEntries" : 1,
  "currentLedgerSize" : 825,
  "lastLedgerCreatedTimestamp" : "2020-07-31T01:47:53.954Z",
  "waitingCursorsCount" : 1,
  "pendingAddEntriesCount" : 0,
  "lastConfirmedEntry" : "491:0",
  "state" : "LedgerOpened",
  "ledgers" : [ {
    "ledgerId" : 448,
    "entries" : 5,
    "size" : 4133,
    "offloaded" : false
  }, {
    "ledgerId" : 456,
    "entries" : 5,
    "size" : 4134,
    "offloaded" : false
  }, {
    "ledgerId" : 464,
    "entries" : 5,
    "size" : 4134,
    "offloaded" : false
  }, {
    "ledgerId" : 471,
    "entries" : 5,
    "size" : 4132,
    "offloaded" : false
  }, {
    "ledgerId" : 482,
    "entries" : 5,
    "size" : 4132,
    "offloaded" : false
  }, {
    "ledgerId" : 491,
    "entries" : 0,
    "size" : 0,
    "offloaded" : false
  } ],
  "cursors" : {
    "normal-subscription" : {
      "markDeletePosition" : "448:3",
      "readPosition" : "491:1",
      "waitingReadOp" : true,
      "pendingReadOps" : 0,
      "messagesConsumedCounter" : 703,
      "cursorLedger" : 486,
      "cursorLedgerLastEntry" : 2,
      "individuallyDeletedMessages" : 
"[(456:-1..456:4],(464:-1..464:4],(471:-1..471:4],(482:-1..482:4],(491:-1..491:0]]",
      "lastLedgerSwitchTimestamp" : "2020-07-30T23:43:50.163Z",
      "state" : "Open",
      "numberOfEntriesSinceFirstNotAckedMessage" : 23,
      "totalNonContiguousDeletedMessagesRange" : 5,
      "properties" : { }
    }
  }
} ```
----
2020-07-31 03:57:23 UTC - Takahiro Hozumi: &gt; It is also has additional 
overhead
Depending on the degree of overhead, I think that deduplication by default 
would be a cost-effective setting. I could think of much use cases where 
duplication is harmful rather than acceptable.

Thank you for your answer.
----
2020-07-31 04:22:17 UTC - Sijie Guo: ```individuallyDeletedMessages" : 
"[(456:-1..456:4],(464:-1..464:4],(471:-1..471:4],(482:-1..482:4],(491:-1..491:0]]",```
This means some of the acknowledgements are missing.
----
2020-07-31 04:25:28 UTC - Sijie Guo: Have you set ackTimeout?
----

Reply via email to