Fix a broken sstable. The scrub process rewrites the sstable, skipping any corrupted rows. Because these rows are lost, follow this process with a repair.
Cassandra must be stopped before this tool is executed, or unexpected results will occur. Note: the script does not verify that Cassandra is stopped.
sstablescrub <options> <keyspace> <table>
|–debug||display stack traces|
Option whether and how to perform a check of the sstable serialization-headers and fix , fixable issues. Possible argument values:
|-h,–help||display this help message|
|-m,–manifest-check||only check and repair the leveled manifest, without actually scrubbing the sstables|
|-n,–no-validate||do not validate columns using column validator|
|-r,–reinsert-overflowed-ttl||Rewrites rows with overflowed expiration date affected by CASSANDRA-14092 with the maximum supported expiration date of 2038-01-19T03:14:06+00:00. The rows are rewritten with the original timestamp incremented by one millisecond to override/supersede any potential tombstone that may have been generated during compaction of the affected rows.|
|-s,–skip-corrupted||skip corrupt rows in counter tables|
The scrub without options will do a snapshot first, then write all non-corrupted files to a new sstable.
sstablescrub keyspace1 standard1 Pre-scrub sstables snapshotted into snapshot pre-scrub-1534424070883 Scrubbing BigTableReader(path='/var/lib/cassandra/data/keyspace1/standard1-6365332094dd11e88f324f9c503e4753/mc-5-big-Data.db') (17.142MiB) Scrub of BigTableReader(path='/var/lib/cassandra/data/keyspace1/standard1-6365332094dd11e88f324f9c503e4753/mc-5-big-Data.db') complete: 73367 rows in new sstable and 0 empty (tombstoned) rows dropped Checking leveled manifest
Use the –no-validate option to retain data that may be misrepresented (e.g., an integer stored in a long field) but not corrupt. This data usually doesn not present any errors to the client.
sstablescrub --no-validate keyspace1 standard1 Pre-scrub sstables snapshotted into snapshot pre-scrub-1536243158517 Scrubbing BigTableReader(path='/var/lib/cassandra/data/keyspace1/standard1-bc9cf530b1da11e886c66d2c86545d91/mc-2-big-Data.db') (4.482MiB) Scrub of BigTableReader(path='/var/lib/cassandra/data/keyspace1/standard1-bc9cf530b1da11e886c66d2c86545d91/mc-2-big-Data.db') complete; looks like all 0 rows were tombstoned
If counter tables are corrupted in a way that prevents sstablescrub from completing, you can use the –skip-corrupted option to skip scrubbing those counter tables. This workaround is not necessary in versions 2.0+.
sstablescrub --skip-corrupted keyspace1 counter1
Using the option –reinsert-overflowed-ttl allows a rewriting of rows that had a max TTL going over the maximum (causing an overflow).
sstablescrub --reinsert-overflowed-ttl keyspace1 counter1
As of Cassandra version 2.0, this option is no longer relevant, since level data was moved from a separate manifest into the sstable metadata.