Pulsar: URI("pulsar://broker-:6650").getPort() will return "-1" instead of 6650

[zhaijack]

found this by wrongly set hostname to "broker-", while using docker to do some test, not sure if this is a bug in java.net.URI.
while after change hostname to "broker-1", it worked fine to get 6650.

Expected behavior

In this part of code,

   public BrokerService(PulsarService pulsar) throws Exception {
        this.pulsar = pulsar;
        this.managedLedgerFactory = pulsar.getManagedLedgerFactory();
        this.port = new URI(pulsar.getBrokerServiceUrl()).getPort();
        this.tlsPort = new URI(pulsar.getBrokerServiceUrlTls()).getPort();
...

URI("pulsar://broker-:6650").getPort() should return 6650

Actual behavior

URI("pulsar://broker-:6650").getPort() will return "-1"

Steps to reproduce

set hostname to "broker-", start broker on this host, it will report wrong port:

2018-01-03 02:13:14,778 - INFO  - [main:DistributedIdGenerator@79] - Created sequential node at /counters/producer-name/-0000000000 -- Generator Id is test_zj-0
2018-01-03 02:13:14,781 - ERROR - [main:PulsarService@345] - port out of range:-1
java.lang.IllegalArgumentException: port out of range:-1
    at java.net.InetSocketAddress.checkPort(InetSocketAddress.java:143)
    at java.net.InetSocketAddress.<init>(InetSocketAddress.java:224)
    at org.apache.pulsar.broker.service.BrokerService.start(BrokerService.java:285)
    at org.apache.pulsar.broker.PulsarService.start(PulsarService.java:270)
    at org.apache.pulsar.PulsarBrokerStarter$BrokerStarter.start(PulsarBrokerStarter.java:190)

the docker command that I wrongly set hostname is this:

docker run -it\
        --network pulsar_network \
        --volume /tmp/test_bk/broker-/journal:/data/journal \
        --volume /tmp/test_bk/broker-/ledger:/data/ledger \
        --volume /tmp/test_bk/broker-/index:/data/index \
        --name "broker-" \
        --hostname "broker-" \
        --env zkServers=192.168.123.71:2181 \
        --env zookeeperServers=192.168.123.71:2181 \
        --env globalZookeeperServers=192.168.123.71:2181 \
        --env clusterName=test_zhaijia \
        --env statsProviderClass=org.apache.bookkeeper.stats.NullStatsProvider \
        --env zkLedgersRootPath=/ledgers \
            bin/apply-config-from-env.py conf/broker.conf &&  bin/apply-config-from-env.py conf/pulsar_env.sh && bin/apply-config-from-env.py conf/bookkeeper.conf  && bin/pulsar broker

[sijie]

as discussed, we might consider documenting what special characters can't be used in pulsar service uri.

[sijie]

let's try improve error handling and produce meaningful message if we can't get around with the java URL api.

[sijie]

@zhaijack can you work on this? since it has been there for a while. it should be a simple fix.

[ppgoml]

It seems I cannot use '_' in host.
(new URI("pulsar://tkiot_pulsar1:6650")).getPort() returns -1

I found this useful link.
https://bugs.java.com/bugdatabase/view_bug.do?bug_id=5049974

[sijie]

@ppgoml yes it is a problem from java's URI

[sijie]

Closed this issue since it is a problem in the URI.