2018-06-15 11:08:19 UTC - Idan: Hi, i would like to know if there is 
possibility via api to purge queue ?
----
2018-06-15 11:08:38 UTC - Idan: (java client)
----
2018-06-15 11:19:29 UTC - Matti-Pekka Laaksonen: With 2.0, the consumers can 
now listen to multiple topics. Is there any way to access the topic the 
specific message was written to?
----
2018-06-15 11:20:01 UTC - Matti-Pekka Laaksonen: Or is the preferred way to 
distinguish between different types of messages to use message properties?
----
2018-06-15 11:20:37 UTC - Ivan Kelly: @Idan as in drop all messages in a topic?
----
2018-06-15 11:21:14 UTC - Idan: @Ivan Kelly i guess yes
----
2018-06-15 11:21:19 UTC - Idan: delete all messages in topic
----
2018-06-15 11:21:46 UTC - Ivan Kelly: for one subscription, or all 
subscriptions?
----
2018-06-15 11:24:37 UTC - Ivan Kelly: in any case, you want either 
<http://pulsar.apache.org/api/admin/org/apache/pulsar/client/admin/Namespaces.html#clearNamespaceBacklog-java.lang.String->
 or 
<http://pulsar.apache.org/api/admin/org/apache/pulsar/client/admin/Topics.html#expireMessagesForAllSubscriptions-java.lang.String-long->
----
2018-06-15 11:25:16 UTC - Ivan Kelly: @Matti-Pekka Laaksonen the data is there 
for sure, one sec
----
2018-06-15 11:28:50 UTC - Ivan Kelly: @Matti-Pekka Laaksonen you need to cast 
the messageId to a org.apache.pulsar.client.impl.TopicMessageIdImpl and call 
getTopicName() on that
----
2018-06-15 11:29:40 UTC - Matti-Pekka Laaksonen: @Ivan Kelly Oh, nice. This is 
exactly what I was looking for. Thank you!
----
2018-06-15 11:31:26 UTC - Idan: @Ivan Kelly one subscription
----
2018-06-15 11:31:38 UTC - Idan: all topics in that subscription
----
2018-06-15 11:31:54 UTC - Idan: ok got it. thanks!
----
2018-06-15 11:32:05 UTC - Ivan Kelly: 
<http://pulsar.apache.org/api/admin/org/apache/pulsar/client/admin/Namespaces.html#clearNamespaceBacklogForSubscription-java.lang.String-java.lang.String->
----
2018-06-15 11:32:10 UTC - Ivan Kelly: :thumbsup:
----
2018-06-15 15:31:46 UTC - Matteo Merli: @Idan In addition to the Java admin lib 
(and REST) it’s possible to perform these operations through the CLI tool. eg :

```
bin/pulsar-admin topics clear-backlog my-topic -s my-subscription
```
----
2018-06-15 16:01:33 UTC - Idan: btw: how can I find out which namesame my topic 
is using? (using this api: 
<http://pulsar.apache.org/api/admin/org/apache/pulsar/client/admin/Namespaces.html#clearNamespaceBacklogForSubscription-java.lang.String-java.lang.String->)
----
2018-06-15 16:01:56 UTC - Idan: when I consume i never provided namespace only 
topic name and subscription
----
2018-06-15 16:02:33 UTC - Idan: consumer = client.newConsumer()
                    .topic(topic)
                    .subscriptionType(subscriptionType)
                    .ackTimeout(ackTimeoutInMillis, TimeUnit.MILLISECONDS)
                    .subscriptionName(subscriptionName)
                    .subscribe();
----
2018-06-15 16:02:49 UTC - Matteo Merli: a topic is tied to a namespace, the 
topic name already includes the namespace. eg: 

Topic name: `<persistent://my-tenant/my-namespace/my-topic>`
Namespace: `my-tenant/my-namespace`
----
2018-06-15 16:03:13 UTC - Idan: ok i must admit i didnt understand well enough 
the tenant/namspace subjects. i better go the docs again
----
2018-06-15 16:03:14 UTC - Idan: to get this clear
----
2018-06-15 16:03:41 UTC - Idan: (not sure when I set my tenant)
----
2018-06-15 16:04:42 UTC - Matteo Merli: if you use the short topic names. eg 
`my-topic`, that is equivalent to use `public/default/my-topic` (or 
`<persistent://public/default/my-topic`>)
----
2018-06-15 16:05:12 UTC - Idan: to tenant name in this case would be default?
----
2018-06-15 16:05:39 UTC - Idan: public
----
2018-06-15 16:05:50 UTC - Matteo Merli: tenant is `public` — namespace is 
`public/default`
----
2018-06-15 16:06:09 UTC - Idan: why would i want multiple tenants. what 
advantages it can gain
----
2018-06-15 16:06:16 UTC - Idan: i understand you have tenant and namesapces
----
2018-06-15 16:06:24 UTC - Idan: where namespaces relayed under tenants
----
2018-06-15 16:07:40 UTC - Matteo Merli: The idea is that you can have multiple 
tenants in a single cluster and each tenants is allowed to “administer itself”, 
create namespace, grant permissions and perform operations (like clear backlog 
and similar), without the system administrator being involved
----
2018-06-15 16:08:23 UTC - Idan: ok so it will be better to seperate for example 
my production env and dev env into diff tenants
----
2018-06-15 16:08:40 UTC - Idan: like prod-tenant and dev-tenant where i can set 
diff permissions to each
----
2018-06-15 16:08:41 UTC - Idan: true?
----
2018-06-15 16:08:47 UTC - Matteo Merli: yes, that is an good example
----
2018-06-15 16:08:51 UTC - Idan: (couldnt think about other case)
----
2018-06-15 16:09:03 UTC - Idan: what about namespaces.. why would I need 
multiple namespaces?
----
2018-06-15 16:09:38 UTC - Matteo Merli: for example, in places like Yahoo, 
there 80-100 different tenants on a Pulsar cluster
----
2018-06-15 16:09:53 UTC - Matteo Merli: each tenant is a “team” within the 
company
----
2018-06-15 16:11:20 UTC - Matteo Merli: each team will either have different 
applications that use Pulsar, or in any case they will have different 
environments (prod, staging, dev). For all these usages, they will have a 
different namespace, within their team specific tenant
----
2018-06-15 16:12:03 UTC - Idan: ok.. so in my case.. we are new startup we dont 
have lots of teams if at at.. we do have 3 envs (dev,stg,prod) a setup could be 
like that:
----
2018-06-15 16:12:10 UTC - Idan: one tenant with 3 namespaces (for each env)
----
2018-06-15 16:12:21 UTC - Idan: and ofcourse the same topics under each 
namespace
----
2018-06-15 16:13:15 UTC - Matteo Merli: yes, in your case there’s no practical 
difference between using 1-tenant 3-namespaces vs 3-tenants 1-namespace-each
----
2018-06-15 16:15:19 UTC - Idan: i understand
----
2018-06-15 16:15:33 UTC - Idan: ok. btw: i finished to create our dlq 
mechanisem perhaps you want to review my architecture?
----
2018-06-15 16:15:38 UTC - Idan: pretty simple
----
2018-06-15 16:16:05 UTC - Idan: there is one thing iam worried about.. we cant 
make sure our consumers wont consume dedups
----
2018-06-15 16:16:50 UTC - Idan: if pulsar will have any glitch and the consumer 
will consume the same messageId the same.. it will pass our mechanism
----
2018-06-15 16:22:02 UTC - Matteo Merli: @Idan please share details around your 
DLQ implementation
----
2018-06-15 16:31:24 UTC - Idan: ok, so we created shared lib among our 
consumers. the lib has redis client (assuming each of our services has redis). 
when we ‘gather’ the messages from pulsar we first determine it’s 
recieveCounter val by matching the key of messageID. then we update it on redis 
and return it (we using redi’s MULTI command to incr and expire the key). now 
we have another method that checks if the message shouldDispatchToDLQ. in case 
the retry=X (our case is 3) we will automaticlly dispatch the message to a DLQ 
topic by the convention &lt;topic-name&gt;-DLQ
----
2018-06-15 16:31:42 UTC - Idan: the dlq has a consumer that will consume and 
put it somewhere
----
2018-06-15 16:32:47 UTC - Idan: the only caveat here is that if the same 
message will be consume by pulsar-client we will consider it as re-try and let 
it pass through
----
2018-06-15 16:33:13 UTC - Idan: we cant have both re-try determinator and 
dedup-checker (until you guys will send us back the retrycounter from the broker
----
2018-06-15 17:59:08 UTC - Karthik Palanivelu: @Karthik Palanivelu uploaded a 
file: 
<https://apache-pulsar.slack.com/files/U7VRE0Q1G/FB8L5TX29/-.txt|Untitled> and 
commented: @Sijie Guo @Ali Ahmed Proxy is starting only pulsar in 6650 and 
8080. I am using TLS and is not listening on that port. Above is my conf file. 
Do I need to update discovery.conf or Client.conf. For starting proxy, I am 
only configuring proxy.conf?
----
2018-06-15 18:08:56 UTC - Karthik Palanivelu: @Karthik Palanivelu uploaded a 
file: <https://apache-pulsar.slack.com/files/U7VRE0Q1G/FB9HJFSTZ/-.sh|Untitled> 
and commented: Within the Proxy Container
----
2018-06-15 18:19:01 UTC - Sijie Guo: @Karthik Palanivelu: # Enable TLS in the 
proxy
tlsEnabledInProxy=false
----
2018-06-15 18:19:07 UTC - Sijie Guo: I thik you need to enable tls
----
2018-06-15 18:36:07 UTC - Eric Henry: @Eric Henry has joined the channel
----
2018-06-15 18:37:05 UTC - Eric Henry: Hi I'm setting up a new cluster and 
getting a ServiceUnitNotReadyException...how can I fix this?
----
2018-06-15 18:52:27 UTC - Sanjeev Kulkarni: @Eric Henry is this standalone 
cluster? Whats your environment?
----
2018-06-15 18:54:06 UTC - Eric Henry: yep, its a standalone cluster. Using 
incubator version 1.21.0. I have a three node cluster and a three node 
zookeeper cluster. Everything came up fine except I keep getting "Namespace not 
served by this instance"
----
2018-06-15 18:56:02 UTC - Eric Henry: or rather a "single cluster"...deployed 
using the instructions here: 
<https://pulsar.incubator.apache.org/docs/latest/deployment/cluster/>
----
2018-06-15 19:01:44 UTC - Eric Henry: @Sanjeev Kulkarni What does this error 
mean?
----
2018-06-15 19:06:34 UTC - Sanjeev Kulkarni: @Eric Henry any specific reason for 
using 1.21.0? Any reason not to use 1.22 or even better 2.0?
----
2018-06-15 19:07:29 UTC - Eric Henry: @Sanjeev Kulkarni It's just the version 
we have deployed currently. We've been using it in production for several 
months....first time I've encountered this error setting up a new cluster.
----
2018-06-15 20:26:11 UTC - Sijie Guo: @Eric Henry it is probably related to a 
namespace not being owned by one worker and the request being redirected. do 
you have any logs at the server side?
----
2018-06-15 21:17:24 UTC - Matteo Merli: @Eric Henry Apart from the 
recommendation upgrade to 1.22 (which fixes several issues) or 2.0. Can you 
past the exception stack trace? Also it would help if you can describe which 
step of the deployment and when it happens.
----
2018-06-15 22:03:30 UTC - Eric Henry: @Matteo Merli @Sijie Guo hey thanks so 
much for responding...I found my issue. I had a typo in my cluster name when I 
created the metadata in zookeeper. Consequently no instances were serving a 
namespace with that cluster name. But I'm definitely taking your recommendation 
and getting my version bumped to 2.0 during my next maintenance window
+1 : Sijie Guo, Matteo Merli
----
2018-06-16 07:07:48 UTC - Idan: @Ivan Kelly but what if I dont have admin 
installed. using only java client. no way to drop everything via that? i see 
this api is only for the admin app
----

Reply via email to