We have tossed around the idea starting some "Beam meetups" -- maybe now that things are really picking up we should get one off the ground!
On Sat, Oct 1, 2016 at 10:15 PM, Amir Bahmanyari <[email protected]> wrote: > Hi Sumit > I diagnosed it > You are right Kafka not accessible > I increased ulimit to an astronomical number and passed the issue > Maybe we should meet someday and share experiences > I live in the Bay Area > Cheers > Amir > > > Sent from my iPhone > > On Oct 1, 2016, at 10:10 PM, Chawla,Sumit <[email protected]> wrote: > > I have seen this error when Kafka is not accessible. Can you check > connectivity to kafka? Can you read messages from Kafka using console > consumer from the machine where you are running the flink job? > > Regards > Sumit Chawla > > > On Sat, Oct 1, 2016 at 12:40 PM, amir bahmanyari <[email protected]> > wrote: > >> Hi Stephan et al. >> I changed ulimit -n at all servers: Flink & Kafka to 10000 & 16000 >> respectively. >> Rebooted. Retried. Failed the same. >> Then restarted everything from the scratch with disk clean up etc. >> Started fine & no more Timeout due to not being able to fetch the topic >> metadata. >> All that started when I increased the deg of parallelism & created a >> Kafka topic with that number for partitions: 2048. >> Its running now. Am hoping to get better perf due to increasing deg >> paralleism. >> Have a great weekend & Appreciate your feedback... >> Amir- >> >> >> ------------------------------ >> *From:* amir bahmanyari <[email protected]> >> *To:* "[email protected]" <[email protected]> >> *Sent:* Friday, September 30, 2016 2:36 PM >> >> *Subject:* Re: KafkaIO() Timeout expired while fetching kafka topic >> metadata >> >> Hi Stephan, >> yes, this program uses Beam's KafkaIO(). >> >> PCollection<KV<String, String>> kafkarecords = p >> .apply(*KafkaIO.read()*.withBootstrapServers("kafka01:9092"). >> withTopics(topics) >> .withValueCoder(StringUtf8Coder.of()).withoutMetadata()) >> .apply("startBundle", ParDo.of( >> new DoFn<KV<byte[], String>, KV<String, String>>() { >> >> Thanks+regards, >> Amir- >> >> >> ------------------------------ >> *From:* Stephan Ewen <[email protected]> >> *To:* [email protected]; amir bahmanyari < >> [email protected]> >> *Sent:* Friday, September 30, 2016 2:13 PM >> *Subject:* Re: KafkaIO() Timeout expired while fetching kafka topic >> metadata >> >> Not sure if this is caused by the Flink Kafka Consumer (or if this >> program uses Beam's Kafka IO). >> >> If it is using Flink, there is a requirement that the client that submits >> the job can access Kafka. If it cannot, that may be a cause for such a >> timeout. >> >> >> On Fri, Sep 30, 2016 at 7:11 PM, amir bahmanyari <[email protected]> >> wrote: >> >> No more stack trace Raghu. >> Keeping the configs like I described, yes it fails like this since for >> some reason the Kafka Zookeeper gets killed. >> Thanks +regards >> Amir- >> >> ------------------------------ >> *From:* Raghu Angadi <[email protected]> >> *To:* [email protected] ; amir bahmanyari < >> [email protected]> >> *Sent:* Thursday, September 29, 2016 11:28 PM >> *Subject:* Re: KafkaIO() Timeout expired while fetching kafka topic >> metadata >> >> >> On Thu, Sep 29, 2016 at 9:34 PM, amir bahmanyari <[email protected]> >> wrote: >> >> at org.apache.flink.client. CliFrontend.parseParameters( >> CliFrontend.java:1189) >> at org.apache.flink.client. CliFrontend.main(CliFrontend. >> java:1239) >> Caused by: o*rg.apache.kafka.common. errors.TimeoutException: Timeout >> expired while fetching topic metadata* >> >> >> Is there any more stacktrace printed after this line? >> >> Does it always fail? >> >> >> >> >> >> >> >> >> >
