[KAFKA-17792] header parsing ends up timing out and using large quantities of memory if the string looks like a number

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Component/s: connect
• Assignee: Unassigned
• Reporter: Martin Sillence
• Created: 14/Oct/24 17:26
• Updated: 15/Oct/24 14:02

Description

We have trace headers such as:

"X-B3-SpanId": "74320e6e26adc8f8"

if however the value happens to be: "407127e212797209"

This is then treated as a numeric value and it tries to convert this as a numeric representation and an exact value using BigDecimal

we end up with the trace:

    BigDecimal.setScale(int, RoundingMode) line: 2876    
    Values$ValueParser.parseAsExactDecimal(BigDecimal) line: 1044    
    Values$ValueParser.parseAsNumber(String) line: 1025    
    Values$ValueParser.parseNextToken(boolean, String) line: 892    
    Values$ValueParser.parse(boolean) line: 875    
    Values.parseString(String) line: 415    
    SimpleHeaderConverter.toConnectHeader(String, String, byte[]) line: 68    
    WorkerSinkTask.convertHeadersFor(ConsumerRecord<byte[],byte[]>) line: 578    

this takes a long time to convert to an exact representation of a 212 billion digit integer