cassandra-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Nimi Wariboko Jr <>
Subject NullPointerException when trying to compact under 3.2
Date Mon, 25 Jan 2016 23:58:32 GMT

I recently upgraded from 2.1.12 to 3.2, and one issue I'm having is I can
no longer read certain rows from a table. A simple SELECT * FROM `table`
times out, only when the bad partition keys are reached. Trying to query
the affected partition keys directly also causes a timeout.

I think the SSTable might be corrupted because nodetool compact (and
repair) fails (although scrub succeeds). With Debug logging, a compact
results in:

WARN  [CompactionExecutor:5] 2016-01-25 15:13:02,734
- Reading cardinality from Statistics.db failed for
ERROR [CompactionExecutor:5] 2016-01-25 15:13:02,777 - Exception in thread
java.lang.NullPointerException: null

(the NPE has no stack trace).

On disk, the SSTable "ma-3778-big-Data.db" does not exist. Even if I scrub
or restart, Cassandra seems to always try to compact this non existent
table (the SSTables on Disk are numbered 3851+).

I'm assuming the SSTables are fubared, and I'd like to restore a snapshot,
but the fact this "ghost" sstable causes the compaction to fail, I'm unsure
if restoring a new set of sstables would actually solve the issue.


View raw message