[
https://issues.apache.org/jira/browse/CASSANDRA-20570?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
David Capwell updated CASSANDRA-20570:
--------------------------------------
Change Category: Operability
Complexity: Low Hanging Fruit
Status: Open (was: Triage Needed)
> Leveled Compaction doesn't validate maxBytesForLevel when the table is
> altered/created
> --------------------------------------------------------------------------------------
>
> Key: CASSANDRA-20570
> URL: https://issues.apache.org/jira/browse/CASSANDRA-20570
> Project: Apache Cassandra
> Issue Type: Improvement
> Components: Local/Compaction
> Reporter: David Capwell
> Priority: Normal
>
> In fuzz testing I hit this fun error message
> {code}
> java.lang.RuntimeException: Repair job has failed with the error message:
> Repair command #1 failed with error At most 9223372036854775807 bytes may be
> in a compaction level; your maxSSTableSize must be absurdly high to compute
> 4.915501902751334E24. Check the logs on the repair participants for further
> details
> at
> org.apache.cassandra.tools.RepairRunner.progress(RepairRunner.java:187)
> {code}
> I was able to create a table, write data to it, and it only ever had issues
> once I did a repair…
> The error comes from
> {code}
> INFO [node2_Repair-Task:1] 2025-04-18 12:08:14,920 SubstituteLogger.java:169
> - ERROR 19:08:14 Repair 138143d4-1dd2-11b2-ba62-f56ce069092d failed:
> java.lang.RuntimeException: At most 9223372036854775807 bytes may be in a
> compaction level; your maxSSTableSize must be absurdly high to compute
> 4.915501902751334E24
> at
> org.apache.cassandra.db.compaction.LeveledManifest.maxBytesForLevel(LeveledManifest.java:191)
> at
> org.apache.cassandra.db.compaction.LeveledManifest.maxBytesForLevel(LeveledManifest.java:182)
> at
> org.apache.cassandra.db.compaction.LeveledManifest.getEstimatedTasks(LeveledManifest.java:651)
> at
> org.apache.cassandra.db.compaction.LeveledManifest.getEstimatedTasks(LeveledManifest.java:640)
> at
> org.apache.cassandra.db.compaction.LeveledManifest.getEstimatedTasks(LeveledManifest.java:635)
> at
> org.apache.cassandra.db.compaction.LeveledCompactionStrategy.getEstimatedRemainingTasks(LeveledCompactionStrategy.java:276)
> at
> org.apache.cassandra.db.compaction.CompactionStrategyManager.getEstimatedRemainingTasks(CompactionStrategyManager.java:1147)
> at
> org.apache.cassandra.metrics.CompactionMetrics$1.getValue(CompactionMetrics.java:81)
> at
> org.apache.cassandra.metrics.CompactionMetrics$1.getValue(CompactionMetrics.java:73)
> at
> org.apache.cassandra.db.compaction.CompactionManager.getPendingTasks(CompactionManager.java:2368)
> at
> org.apache.cassandra.service.ActiveRepairService.verifyCompactionsPendingThreshold(ActiveRepairService.java:659)
> at
> org.apache.cassandra.service.ActiveRepairService.prepareForRepair(ActiveRepairService.java:672)
> at
> org.apache.cassandra.repair.RepairCoordinator.prepare(RepairCoordinator.java:486)
> at
> org.apache.cassandra.repair.RepairCoordinator.runMayThrow(RepairCoordinator.java:329)
> at
> org.apache.cassandra.repair.RepairCoordinator.run(RepairCoordinator.java:280)
> {code}
> Which has this logic
> {code}
> double bytes = Math.pow(levelFanoutSize, level) * maxSSTableSizeInBytes;
> if (bytes > Long.MAX_VALUE)
> throw new RuntimeException("At most " + Long.MAX_VALUE + " bytes may be
> in a compaction level; your maxSSTableSize must be absurdly high to compute "
> + bytes);
> {code}
> The fuzz test had the following inputs
> {code}
> level = 8
> levelFanoutSize = 90
> maxSSTableSizeInBytes = 1141899264
> {code}
> Given that the max level is known (its 8, and is hard coded), we can do this
> calculation during create table / alter table to make sure that we don’t blow
> up later on
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org
For additional commands, e-mail: commits-h...@cassandra.apache.org
