2020-06-25 09:31:54 UTC - prem m: @prem m has joined the channel
----
2020-06-25 09:36:09 UTC - prem m: Hi All, may I know from pulsar functions how 
to publish message(s)  to different pulsar topic partition based on hashing? . 
pulsar producer has various hashing schemes, like that i am looking for.
----
2020-06-25 12:44:49 UTC - Rahul Vashishth: @sjmittal @Vladimir Shchur 
@dionjansen same issue for me too. did you find a fix or workaround?
----
2020-06-25 12:45:15 UTC - Vladimir Shchur: Workaround is to rollback to 2.5.2
+1 : Rahul Vashishth
----
2020-06-25 13:12:32 UTC - Rahul Vashishth: Even after 
<https://github.com/apache/pulsar-helm-chart/blob/master/charts/pulsar/values.yaml#L113|dashboard:
 false> in charts, but still i can see dashboard pods being created when i 
installed latest chart 2.6.0
----
2020-06-25 13:49:29 UTC - Matteo Merli: No, the termination is permanently 
sealing the topic
----
2020-06-25 13:49:45 UTC - Jeff Schneller: We have a cluster of 3 brokers and 
trying to setup pulsar-manager to manage it.  However pulsar-manager will not 
connect to the cluster.  The 3 brokers have authentication and authorization 
turned on.  For Pulsar Manager we created a keystore with our ca and the cert 
for the admin role.  However it will not connect.    Any ideas?  What url do we 
use for the cluster environment?   How is the cluster defined?  We are able to 
run the admin-cli from one of the brokers without any issues.
----
2020-06-25 14:07:58 UTC - Frank Kelly: Thanks for flagging this - we are 
planning a custom auth methodology that uses a new JWT format - hopefully we'll 
just be able to piggy back on the current Auth Token provider (wrapping a JWT 
token with different semantics). So this is good to know
----
2020-06-25 14:22:58 UTC - Gavin Liu: Have you tried just connecting to one 
broker instead of 3, I know that for version 2.5.1 and 2.5.2, if the pulsar 
connection string contains more than 1 broker, you have trouble accessing some 
functionalities.
----
2020-06-25 14:24:03 UTC - Jeff Schneller: using only 1.  the manager connects 
to  http/https like the admin cli not the pulsar ports.
+1 : Gavin Liu
----
2020-06-25 14:25:15 UTC - Gavin Liu: @Poul Henriksen Hey man, I searched the 
history and found that you asked about this as well, did you figure all out?
----
2020-06-25 14:43:10 UTC - Poul Henriksen: no...
----
2020-06-25 15:00:29 UTC - Charvak Patel: in the url, are you giving tcp port ( 
mostly 8080 )?
you can see this as an example setup for pulsar manager
<https://pulsar.apache.org/docs/en/kubernetes-helm/#step-4-use-pulsar-manager-to-manage-the-cluster>
The url provided is <http://pulsar-mini-broker:8080>
which is this
NAME                         TYPE           CLUSTER-IP       EXTERNAL-IP   
PORT(S)                       AGE
pulsar-mini-bookie           ClusterIP      None             &lt;none&gt;       
 3181/TCP,8000/TCP             11m
pulsar-mini-broker           ClusterIP      None             &lt;none&gt;       
 8080/TCP,6650/TCP             11m
----
2020-06-25 15:04:08 UTC - Gavin Liu: Bummer, I am almost affirmative that all 
these are on a partition level. Because if not, when over the limit, how the 
events are are deleted from the partitions?  @Sijie Guo agree?
----
2020-06-25 15:11:11 UTC - Jeff Schneller: not using kubernetes and 8080 
probably won't work since we are using tls certs
----
2020-06-25 15:15:36 UTC - Anup Ghatage: Hey @rani
I wrote a doc about how to cleanly decomm bookies.

Check it out here:
<https://pulsar.apache.org/docs/en/administration-zk-bk/#decommissioning-bookies-cleanly>
----
2020-06-25 15:40:09 UTC - Rahul Vashishth: it seems similar issue was fixed in 
<https://github.com/apache/pulsar/issues/6355> and reintroduced again
----
2020-06-25 17:41:35 UTC - Raphael Enns: I've been doing some local stability 
testing with our messaging tool, which is backed by a pulsar broker. I'm using 
java client 2.5.2 and running pulsar 2.5.2 in standalone mode.
While running a test, suddenly I stopped receiving any messages from pulsar. I 
noticed that the thread "pulsar-external-listener-3-1" in pulsar-client stopped 
having any log entries at that point.
I did a thread dump of the client and it looks like that thread is stuck 
waiting in "hasMessagesAvailable" (see below).
There were no warnings or errors in the client logs or the pulsar broker logs. 
The only way to resolve it is to restart the pulsar client, which is definitely 
not ideal.
----
2020-06-25 17:41:52 UTC - Raphael Enns: 
----
2020-06-25 17:42:52 UTC - Matteo Merli: With tokens, there will be no problem, 
you can use the regular auth provider on client side and a custom one on the 
broker.
sunglasses : Frank Kelly
----
2020-06-25 18:27:00 UTC - Rahul Vashishth: Is the helm chart repo for this 
<https://github.com/apache/pulsar-helm-chart/> is publically hosted somewhere,  
like github pages or helm hub?
----
2020-06-25 18:29:29 UTC - Rahul Vashishth: @Sijie Guo @Adriaan de Haan i see 
the same error on current chart in pulsar-bookkeeper's init container 
(pulsar-bookkeeper-verify-clusterid)

`[conf/bookkeeper.conf] Applying config zkServers = pulsar-zookeeper:2181`
`[conf/bookkeeper.conf] Adding config journalDirectories = 
/pulsar/data/bookkeeper/journal`
`JMX enabled by default`
`Error: Could not find or load main class "-Xms1g`
----
2020-06-25 18:46:33 UTC - Shilpa Shanbhag: @Shilpa Shanbhag has joined the 
channel
----
2020-06-25 20:14:09 UTC - rwaweber: To speak to the same general issue — we 
created another topic, without a consumer, as a means to understand backlog 
retention policy behavior better, and it appears that the `pulsar_storage_size` 
metric is twice the value of what is reported through `./bin/pulsar-admin 
topics stats topic &lt;topicName&gt;`  for the `storageSize` key. What 
calculations go into the `pulsar_storage_size` metric?
----
2020-06-25 22:14:54 UTC - Kirill Merkushev: so even no way to recreate the 
topic?
----
2020-06-25 22:15:48 UTC - Kirill Merkushev: Is there a way to unsubscribe 
exclusive consumer from the broker side?
----
2020-06-25 22:31:36 UTC - Raphael Enns: Another thing I ran into was that I got 
an OutOfMemoryError in the Java pulsar client (2.5.2), which seemed due to 
being disconnected from the broker. Here are a few of the ERROR/WARN log 
entries from the java client:
----
2020-06-25 22:31:52 UTC - Raphael Enns: 
----
2020-06-25 22:32:05 UTC - Raphael Enns: I removed duplicate log entries.
----
2020-06-25 22:32:30 UTC - Matteo Merli: `pulsar-admin topics unsubscribe $TOPIC 
-s $SUB`
+1 : Kirill Merkushev
----
2020-06-25 22:33:07 UTC - Matteo Merli: Yes, you can delete the topic and 
re-create
----
2020-06-25 22:33:14 UTC - Raphael Enns: From what I could gather, the 
OutOfMemoryError occurred because the java client kept accumulating either 
messages to be sent or readers or the data until it ran out of memory.
----
2020-06-25 22:33:25 UTC - Raphael Enns: Do you have any suggestions on how to 
recover from cases like this?
----
2020-06-25 22:34:03 UTC - Matteo Merli: You can control how many messages to 
keep in pending state, with the `producerQueueSize` setting when creating the 
producer.
----
2020-06-25 22:35:16 UTC - Kirill Merkushev: This is odd, since I’ve got an 
error that topic was terminated after I recreated it…ok, anyway thanks for the 
quick response
----
2020-06-25 23:05:39 UTC - Alan Kittel: @Alan Kittel has joined the channel
----
2020-06-25 23:20:13 UTC - Matteo Merli: Ok, if that's the case it would be a 
bug :slightly_smiling_face:
----
2020-06-25 23:30:46 UTC - Sijie Guo: it is per-topic(partition) level.
----
2020-06-25 23:31:09 UTC - Sijie Guo: The namespace policy is applied to the 
topic (partitions) within it.
----
2020-06-25 23:32:56 UTC - Sijie Guo: Are you using tls authentication? Or just 
using tls for encryption?
----
2020-06-25 23:33:29 UTC - Jeff Schneller: Both
----
2020-06-25 23:33:59 UTC - Sijie Guo: Not yet. It will be published to 
<https://pulsar.apache.org/charts>.
----
2020-06-25 23:34:10 UTC - Sijie Guo: There is an outstanding PR to be merged to 
make it happen.
----
2020-06-25 23:37:58 UTC - Kirill Merkushev: its saying that subscription has 
active consumers that way. Is there a force mode?
----
2020-06-25 23:40:37 UTC - Sijie Guo: how did you get the `pulsar_storage_size` ?

• did you curl the metrics endpoint? If so, can you attach the whole metric 
with all the labels?
• did you get the storage_size from grafana? If so, it is worth to see if 
double-counts the metrics from prometheus.
----
2020-06-25 23:44:40 UTC - Sijie Guo: if it is tls authentication, I don’t think 
Pulsar Manager supports tls authentication yet.
----
2020-06-25 23:44:52 UTC - Sijie Guo: It only supports tls encryption + jwt 
authentication.
----
2020-06-25 23:46:18 UTC - Jeff Schneller: Hmm. Ok.  Will need to rethink things 
then. 
----
2020-06-25 23:47:37 UTC - Sijie Guo: Can you create an issue at 
<https://github.com/apache/pulsar-manager>? We can see how to add that support.
----
2020-06-26 00:30:34 UTC - Gavin Liu: @Sijie Guo Thanks for the confirmation. 
:thumbsup: :thumbsup:
----
2020-06-26 01:24:32 UTC - Jeff Schneller: Issue created
ok_hand : Sijie Guo
----
2020-06-26 04:09:18 UTC - Matteo Merli: Yes, just add `--force`
----
2020-06-26 08:38:02 UTC - Sajith Sebastian: @Sijie Guo I ran into an issue with 
pulsar binary protocol when I configure AMQP connector. The moment we configure 
messagingProtocols=amqp default pulsar binary protocol is not listening in the 
port 6650 .  Can you please share some light
----

Reply via email to