Jake Maes commented on SAMZA-1387:

>My problem is with coordinator topic. Maybe there is a difference? Just a 

Ahh, ok now it makes sense, so it's not a topic name that you configured with 
the '.' in the name. Its the job.name property that has a '.' in the name. 

Samza actually doesn't let you configure the coordinator stream topic. It's 
derived from the job name. 

I'll fix that. Thanks for finding it!

In the meantime, is it feasible to change the job.name? Note: this would change 
the names of the checkpoint and coordinator topics, the job would basically 
start from scratch. 

> Unable to Start Samza App Because Regex Check
> ---------------------------------------------
>                 Key: SAMZA-1387
>                 URL: https://issues.apache.org/jira/browse/SAMZA-1387
>             Project: Samza
>          Issue Type: Bug
>    Affects Versions: 0.13.0
>            Reporter: Hao Song
>            Assignee: Jake Maes
> [I created this as a bug but feel free to change it if it's a feature]
> When I'm trying to upgrade Samza to 0.13 (from 0.11), I got the error below 
> when trying to start our job: 
> [ec2-user@namenode-shadow-1] out: 2017-08-09 17:26:31.977 [main] 
> KafkaSystemAdmin [INFO] Attempting to create coordinator stream 
> __samza_coordinator_<env>.<AppName>_1.
> [ec2-user@namenode-shadow-1] out: Exception in thread "main" 
> java.lang.IllegalArgumentException: Identifier 'streamId' is 
> '__samza_coordinator_<env>.<AppName>_1'. It must match the expression 
> [A-Za-z0-9_-]+
> [ec2-user@namenode-shadow-1] out:     at 
> org.apache.samza.system.StreamSpec.validateLogicalIdentifier(StreamSpec.java:201)
> [ec2-user@namenode-shadow-1] out:     at 
> org.apache.samza.system.StreamSpec.<init>(StreamSpec.java:140)
> [ec2-user@namenode-shadow-1] out:     at 
> org.apache.samza.system.kafka.KafkaStreamSpec.<init>(KafkaStreamSpec.java:152)
> [ec2-user@namenode-shadow-1] out:     at 
> org.apache.samza.system.kafka.KafkaSystemAdmin.createCoordinatorStream(KafkaSystemAdmin.scala:334)
> [ec2-user@namenode-shadow-1] out:     at 
> org.apache.samza.job.JobRunner.run(JobRunner.scala:88)
> [ec2-user@namenode-shadow-1] out:     at 
> org.apache.samza.job.JobRunner$.doOperation(JobRunner.scala:52)
> [ec2-user@namenode-shadow-1] out:     at 
> org.apache.samza.job.JobRunner$.main(JobRunner.scala:47)
> [ec2-user@namenode-shadow-1] out:     at 
> org.apache.samza.job.JobRunner.main(JobRunner.scala)
> Looking at Samza code on Github, it seems Samza just started to restrict 
> topic names to a stricter sets. It doesn't allow '.' anymore whereas we have 
> been using '.' for all of our topics. That means we cannot upgrade to Samza 
> 0.13 without changing all the topics. That also breaks our convention for 
> defining topic names. Having '.' in the topic name is actually very 
> convenient for us because they can be shown in Grafana/StatsD in a more 
> organized way.
> Here is the change in Github: 
> https://github.com/apache/samza/commit/e6c1eed4f1d576661abafce8477c1749c2554b39#diff-98a114809fb1ab416f8d1d0b4d318b96R201
> Can you expand the regex set to a larger set including '.'? Thanks in advance!

This message was sent by Atlassian JIRA

Reply via email to