[
https://issues.apache.org/jira/browse/KAFKA-10713?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17231324#comment-17231324
]
Tom Bentley commented on KAFKA-10713:
-------------------------------------
The problem is this regex in {{Utils}}
{noformat}
HOST_PORT_PATTERN =
Pattern.compile(".*?\\[?([0-9a-zA-Z\\-%._:]*)\\]?:([0-9]+)");
{noformat}
The initial {{.*?}}, although a non-greedy quantifier, matches everything upto
and including the last semicolon because the rest of the expression cannot
match sooner.
This pattern is supposed to match
{code}
host:port and protocol:\\host:port
{code}
according to the code comment. [RFC
2396|https://tools.ietf.org/html/rfc2396#section-3.1] says URL schema has to
match {{scheme = alpha *( alpha | digit | "+" | "-" | "." )}}. That's ASCII
alpha. Due to the historic laxity in parsing the protocol/scheme part _perhaps_
we should allow any unicode letter, rather than limiting to ASCII alpha, so
avoid breaking anyone who has been using non-ASCII characters.
> Surprising behaviour when bootstrap servers are separated by semicolons
> -----------------------------------------------------------------------
>
> Key: KAFKA-10713
> URL: https://issues.apache.org/jira/browse/KAFKA-10713
> Project: Kafka
> Issue Type: Improvement
> Reporter: Mickael Maison
> Assignee: Tom Bentley
> Priority: Major
>
> When creating a Kafka client with {{bootstrap.servers}} set to
> "kafka-0:9092;kafka-1:9092;kafka-2:9092", it has a strange behaviour.
> For once, there's no warning or error messages. The client will connect and
> start working. However, it will only use the hostname after the last
> semicolon as bootstrap server!
> The configuration {{bootstrap.servers}} is defined as a {{List}} in
> {{AbstractConfig}}. So from a configuration point of view,
> "kafka-0:9092;kafka-1:9092;kafka-2:9092" is a single entry.
> Then, {{Utils.getHost()}} returns "kafka-2" when parsing that string.
> {code:java}
> assertEquals("kafka-2", getHost("kafka-1:9092;kafka-1:9092;kafka-2:9092"));
> {code}
> So the client ends up with a single bootstrap server!
> I believe semicolon are not valid characters in hostname/domain names, so we
> should be able to provide better validation.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
