2019-09-30 10:49:52 UTC - Vladimir Shchur: @jia zhai So this is how you query state from the function itself, and what I want is to query it from external app, say I have a webserver which exposes getWordCount api. Sorry, I must be missing something ---- 2019-09-30 12:36:30 UTC - Jesse Zhang (Bose): Question, the effective value of `NackRedeliveryDelay` is about 1/4-1/3 of its specified value. Is this a known bug or my usage issue? See my test details in the threads. arrow_up : Jesse Zhang (Bose) ---- 2019-09-30 12:36:43 UTC - Jesse Zhang (Bose): In my test, I set NackRedeliveryDelay=200 second delay, but it is retrying at second 57
The same behavior happened in all these server/client combinations: Standalone Pular server 2.4.1. or 2.3.1 Python or golang, with C++ client 2.4.0 See my log: 1 - ID: (10,0,-1,-1) //at second 1, processed message #0 2 - ID: (10,1,-1,-1) 3 - ID: (10,2,-1,-1) 4 - ID: (10,3,-1,-1) 5 - ID: (10,4,-1,-1) 6 - ID: (10,5,-1,-1) 7 - ID: (10,6,-1,-1) 8 - ID: (10,7,-1,-1) 9 - ID: (10,8,-1,-1) 10 - ID: (10,9,-1,-1) //at second 10, processed message #9 68 - ID: (10,9,-1,-1) //at second 68, processed message #9 again, delay is 57 seconds only. 69 - ID: (10,0,-1,-1) //at second 69, processed message #0 again, delay is 68 seconds only. 70 - ID: (10,1,-1,-1) 71 - ID: (10,2,-1,-1) 72 - ID: (10,3,-1,-1) 73 - ID: (10,4,-1,-1) 74 - ID: (10,5,-1,-1) 75 - ID: (10,6,-1,-1) 76 - ID: (10,7,-1,-1) ---- 2019-09-30 13:36:24 UTC - jia zhai: there are cli, java api and python api to get. ---- 2019-09-30 14:17:27 UTC - Vladimir Shchur: Yes, but from you links I see only that this api can be used within Pulsar functions, not from external app... ---- 2019-09-30 17:16:39 UTC - Rong: @Rong has joined the channel ---- 2019-09-30 17:51:14 UTC - Tarek Shaar: What's best way to view the pay load of messages that are in the backlog for a particular topic? Assuming we identified the ledger that has the message, is the only way to do that is to connect to Book Keeper and use its API? ---- 2019-09-30 18:00:29 UTC - Matteo Merli: `pulsar-admin topics peek $TOPIC -s $SUBSCRIPTION` No need to get it using BookKeeper API (though that’s possible as well) ---- 2019-09-30 18:01:35 UTC - Tarek Shaar: Ok thanks got it ---- 2019-09-30 21:21:32 UTC - Ali Ahmed: I have release 2.4.1 python artifacts for osx <https://pypi.org/project/pulsar-client/2.4.1/#files> hugging_face : Poule, Ali Ahmed ---- 2019-09-30 22:57:38 UTC - Ali Ahmed: I have prepared the vcpkg formulae for cross platform c++ library management <https://github.com/microsoft/vcpkg/pull/5693> ---- 2019-09-30 22:58:29 UTC - Ali Ahmed: ---- 2019-09-30 23:37:18 UTC - Andrey Popelo: Hey guys In Pulsar is it possible to: 1) Publish a message and have it delivered at specific date/time. Or schedule a pulsar function to be executed once at specific date/time. 2) Periodically deliver a message. E.g. every Monday at 01:00 am deliver a message. Or periodically call a pulsar function. ---- 2019-09-30 23:38:21 UTC - Ali Ahmed: @Andrey Popelo have a look here <https://pulsar.apache.org/blog/2019/07/05/Apache-Pulsar-2-4-0/> +1 : Andrey Popelo ---- 2019-10-01 00:54:41 UTC - Addison Higham: huh... so that was weird, we are trying to get the BK table service working for stateful functions and this was our experience: 1. enabled the options in broker/function worker, enabled the table service, it resulted in a loop of JVM segfaults of a single bookie in rocksdb code 2. disabled the component on 2/3 bookies, it stabilized the bookies 3. the function was failing to fetching state, we assume that maybe it had init'ed state on the bookie that was crashing, so we were trying to figure out how to remove the "container" 4. To try and do so, we removed the `streams` tree from ZK 5. this seems to have caused our ZK nodes to get into a failure mode where they kept crashing. Haven't quite grokked the logs yet, but eventually, we just turned off the table service component in our last bookie and the quorum immediately stabilized ---- 2019-10-01 00:55:46 UTC - Addison Higham: does that seem at all plausible? I think we did plenty of things wrong trying to running the table service only on a subset of the bookies, but I am still surprised that it was able to cause my ZK to lose quorum due to crashes of the node ---- 2019-10-01 01:15:35 UTC - Poule: @Vladimir Shchur did you try the rest api ---- 2019-10-01 01:16:19 UTC - Poule: <https://pulsar.apache.org/admin/v3/functions/{tenant}/{namespace}/{functionName}/state/{key}> ---- 2019-10-01 01:18:37 UTC - Poule: not sure about the throughput it can handle though ---- 2019-10-01 01:23:29 UTC - Vladimir Shchur: wow! looks good, thank you! ---- 2019-10-01 01:57:38 UTC - Poule: you do Event Sourcing with aggregates? ---- 2019-10-01 07:06:17 UTC - Vladimir Shchur: Not yet, I'm investigating available options. Having the ability to query aggregates looks very important for me, since no external storage is needed ---- 2019-10-01 07:45:16 UTC - Gaetan SNL: @Gaetan SNL has joined the channel ---- 2019-10-01 07:55:42 UTC - Gaetan SNL: Hello, I discovered Pulsar recently and it seems to solve almost all my needs for message delivery, except one ! I'm not sure if it's a good use case for pulsar. I would like to create temporary subscriptions and if the consumer disapear the subscription is removed, of course I will lost messages when the consumer is disconnected and I'm OK with that. I currently use NATS for this, for things like chat messages, other notifications and message passing between servers. I'm wondering if (1) it's possible and (2) if it's OK to have thousands of susbcription for one consumer. ----
