[CASSANDRA-14672] After deleting data in 3.11.3, reads fail with "open marker and close marker have different deletion times"

• Type: Bug
• Status: Resolved
• Priority: Urgent
• Resolution: Fixed
• Fix Version/s: 3.0.18, 3.11.4, 4.0-alpha1, 4.0
• Component/s: Local/SSTable

Environment

CentOS 7, GCE, 9 nodes, 4TB disk/~2TB full each, level compaction, timeseries data

• Bug Category: Correctness - Recoverable Corruption / Loss
• Severity: Critical
• Complexity: Challenging
• Discovered By: User Report
• Since Version: 3.0.0
• Assignee: Aleksey Yeschenko
• Reporter: Spiros Ioannou

Authors

Aleksey Yeschenko

Reviewers

Blake Eggleston

• Created: 29/Aug/18 07:09
• Updated: 15/May/20 08:01
• Resolved: 25/Sep/18 16:48

Description

We had 3.11.0, then we upgraded to 3.11.3 last week. We routinely perform deletions as the one described below. After upgrading we run the following deletion query:

 

DELETE FROM measurement_events_dbl WHERE measurement_source_id IN ( 9df798a2-6337-11e8-b52b-42010afa015a,  9df7717e-6337-11e8-b52b-42010afa015a, a08b8042-6337-11e8-b52b-42010afa015a, a08e52cc-6337-11e8-b52b-42010afa015a, a08e6654-6337-11e8-b52b-42010afa015a, a08e6104-6337-11e8-b52b-42010afa015a, a08e6c76-6337-11e8-b52b-42010afa015a, a08e5a9c-6337-11e8-b52b-42010afa015a, a08bcc50-6337-11e8-b52b-42010afa015a) AND year IN (2018) AND measurement_time >= '2018-07-19 04:00:00'

 

Immediately after that, trying to read the last value produces an error:

select * FROM measurement_events_dbl WHERE measurement_source_id = a08b8042-6337-11e8-b52b-42010afa015a AND year IN (2018) order by measurement_time desc limit 1;
ReadFailure: Error from server: code=1300 [Replica(s) failed to execute read] message="Operation failed - received 0 responses and 2 failures" info={'failures': 2, 'received_responses': 0, 'required_responses': 1, 'consistency': 'ONE'}

 

And the following exception: 

WARN [ReadStage-4] 2018-08-29 06:59:53,505 AbstractLocalAwareExecutorService.java:167 - Uncaught exception on thread Thread[ReadStage-4,5,main]: {}
java.lang.RuntimeException: java.lang.IllegalStateException: UnfilteredRowIterator for pvpms_mevents.measurement_events_dbl has an illegal RT bounds sequence: open marker and close marker have different deletion times
 at org.apache.cassandra.service.StorageProxy$DroppableRunnable.run(StorageProxy.java:2601) ~[apache-cassandra-3.11.3.jar:3.11.3]
 at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511) ~[na:1.8.0_181]
 at org.apache.cassandra.concurrent.AbstractLocalAwareExecutorService$FutureTask.run(AbstractLocalAwareExecutorService.java:162) ~[apache-cassandra-3.11.3.jar:3.11.3]
 at org.apache.cassandra.concurrent.AbstractLocalAwareExecutorService$LocalSessionFutureTask.run(AbstractLocalAwareExecutorService.java:134) [apache-cassandra-3.11.3.jar:3.11.3]
 at org.apache.cassandra.concurrent.SEPWorker.run(SEPWorker.java:109) [apache-cassandra-3.11.3.jar:3.11.3]
 at java.lang.Thread.run(Thread.java:748) [na:1.8.0_181]
Caused by: java.lang.IllegalStateException: UnfilteredRowIterator for pvpms_mevents.measurement_events_dbl has an illegal RT bounds sequence: open marker and close marker have different deletion times
 at org.apache.cassandra.db.transform.RTBoundValidator$RowsTransformation.ise(RTBoundValidator.java:103) ~[apache-cassandra-3.11.3.jar:3.11.3]
 at org.apache.cassandra.db.transform.RTBoundValidator$RowsTransformation.applyToMarker(RTBoundValidator.java:81) ~[apache-cassandra-3.11.3.jar:3.11.3]
 at org.apache.cassandra.db.transform.BaseRows.hasNext(BaseRows.java:148) ~[apache-cassandra-3.11.3.jar:3.11.3]
 at org.apache.cassandra.db.rows.UnfilteredRowIteratorSerializer.serialize(UnfilteredRowIteratorSerializer.java:136) ~[apache-cassandra-3.11.3.jar:3.11.3]
 at org.apache.cassandra.db.rows.UnfilteredRowIteratorSerializer.serialize(UnfilteredRowIteratorSerializer.java:92) ~[apache-cassandra-3.11.3.jar:3.11.3]
 at org.apache.cassandra.db.rows.UnfilteredRowIteratorSerializer.serialize(UnfilteredRowIteratorSerializer.java:79) ~[apache-cassandra-3.11.3.jar:3.11.3]
 at org.apache.cassandra.db.partitions.UnfilteredPartitionIterators$Serializer.serialize(UnfilteredPartitionIterators.java:308) ~[apache-cassandra-3.11.3.jar:3.11.3]
 at org.apache.cassandra.db.ReadResponse$LocalDataResponse.build(ReadResponse.java:187) ~[apache-cassandra-3.11.3.jar:3.11.3]
 at org.apache.cassandra.db.ReadResponse$LocalDataResponse.<init>(ReadResponse.java:180) ~[apache-cassandra-3.11.3.jar:3.11.3]
 at org.apache.cassandra.db.ReadResponse$LocalDataResponse.<init>(ReadResponse.java:176) ~[apache-cassandra-3.11.3.jar:3.11.3]
 at org.apache.cassandra.db.ReadResponse.createDataResponse(ReadResponse.java:76) ~[apache-cassandra-3.11.3.jar:3.11.3]
 at org.apache.cassandra.db.ReadCommand.createResponse(ReadCommand.java:352) ~[apache-cassandra-3.11.3.jar:3.11.3]
 at org.apache.cassandra.service.StorageProxy$LocalReadRunnable.runMayThrow(StorageProxy.java:1889) ~[apache-cassandra-3.11.3.jar:3.11.3]
 at org.apache.cassandra.service.StorageProxy$DroppableRunnable.run(StorageProxy.java:2597) ~[apache-cassandra-3.11.3.jar:3.11.3]
 ... 5 common frames omitted
 Suppressed: java.lang.IllegalStateException: UnfilteredRowIterator for pvpms_mevents.measurement_events_dbl has an illegal RT bounds sequence: expected all RTs to be closed, but the last one is open
 at org.apache.cassandra.db.transform.RTBoundValidator$RowsTransformation.ise(RTBoundValidator.java:103) ~[apache-cassandra-3.11.3.jar:3.11.3]
 at org.apache.cassandra.db.transform.RTBoundValidator$RowsTransformation.onPartitionClose(RTBoundValidator.java:96) ~[apache-cassandra-3.11.3.jar:3.11.3]
 at org.apache.cassandra.db.transform.BaseRows.runOnClose(BaseRows.java:91) ~[apache-cassandra-3.11.3.jar:3.11.3]
 at org.apache.cassandra.db.transform.BaseIterator.close(BaseIterator.java:86) ~[apache-cassandra-3.11.3.jar:3.11.3]
 at org.apache.cassandra.db.partitions.UnfilteredPartitionIterators$Serializer.serialize(UnfilteredPartitionIterators.java:309) ~[apache-cassandra-3.11.3.jar:3.11.3]
 ... 12 common frames omitted
 

 

We have 9 nodes ~2TB each, leveled compaction, repairs run daily in sequence.

Table definition is:

CREATE TABLE pvpms_mevents.measurement_events_dbl (
 measurement_source_id uuid,
 year int,
 measurement_time timestamp,
 event_reception_time timestamp,
 quality double,
 value double,
 PRIMARY KEY ((measurement_source_id, year), measurement_time)
) WITH CLUSTERING ORDER BY (measurement_time ASC)
 AND bloom_filter_fp_chance = 0.1
 AND caching = {'keys': 'ALL', 'rows_per_partition': 'NONE'}
 AND comment = ''
 AND compaction = {'class': 'org.apache.cassandra.db.compaction.LeveledCompactionStrategy'}
 AND compression = {'chunk_length_in_kb': '64', 'class': 'org.apache.cassandra.io.compress.LZ4Compressor'}
 AND crc_check_chance = 1.0
 AND dclocal_read_repair_chance = 0.1
 AND default_time_to_live = 0
 AND gc_grace_seconds = 864000
 AND max_index_interval = 2048
 AND memtable_flush_period_in_ms = 0
 AND min_index_interval = 128
 AND read_repair_chance = 0.0
 AND speculative_retry = '99PERCENTILE';

 

We host those on GCE and recreated all the nodes with disk snapshots, and we reproduced the error: after re-running the DELETE with all nodes up and no other queries running, the error was reproduced immediately.

 

We tried so far:

re-running repairs on all nodes and running nodetool garbagecollect with no success.

We downgraded to 3.11.2 for now, no issues so far.

This may be related to CASSANDRA-14515

Issue Links

• is caused by: CASSANDRA-14515 Short read protection in presence of almost-purgeable range tombstones may cause permanent data loss