[CASSANDRA-6791] CompressedSequentialWriter can write zero-length segments during scrub

• Type: Bug
• Status: Resolved
• Priority: Low
• Resolution: Fixed
• Fix Version/s: 1.2.16, 2.0.6, 2.1 beta2
• Severity: Low
• Assignee: Viktor Kuzmin
• Reporter: Jonathan Ellis

Authors

Viktor Kuzmin

Reviewers

Marcus Eriksson

• Created: 02/Mar/14 02:18
• Updated: 16/Apr/19 09:31
• Resolved: 03/Mar/14 13:31

Description

This results in errors like this:

java.lang.IllegalArgumentException
	at java.nio.Buffer.limit(Buffer.java:267)
	at org.apache.cassandra.io.compress.CompressedRandomAccessReader.decompressChunk(CompressedRandomAccessReader.java:108)
	at org.apache.cassandra.io.compress.CompressedRandomAccessReader.reBuffer(CompressedRandomAccessReader.java:87)
	at org.apache.cassandra.io.util.RandomAccessReader.seek(RandomAccessReader.java:280)

(Because a zero-length chunk actually turns into a length of -4 in the compressed.limit(chunk.length) call, since no checksum is written either.)

I thought this would be from two bad rows in a row, but it's not; the source file that resulted in scrub creating this, did not have any of those. (But it does have several instances of bad-good-bad, i.e. separated by exactly one row, that is not large enough to force a new compressed chunk.)