[CASSANDRA-9806] some TTL test are failing on trunk: losing data after restart?

• Type: Bug
• Status: Resolved
• Priority: Urgent
• Resolution: Invalid
• Fix Version/s: 3.0.0 rc2
• Component/s: Legacy/Testing
• Severity: Critical
• Assignee: Paulo Motta
• Reporter: Alan Boudreault

Authors

Paulo Motta

• Tester: Alan Boudreault
• Created: 14/Jul/15 19:52
• Updated: 16/Apr/19 09:31
• Resolved: 28/Sep/15 22:03

Description

ttl_test.TestDistributedTTL.ttl_is_respected_on_delayed_replication_test is failing and ttl_test.TestDistributedTTL.ttl_is_respected_on_repair_test are failing:

http://cassci.datastax.com/view/trunk/job/trunk_dtest/346/testReport/junit/ttl_test/TestDistributedTTL/ttl_is_respected_on_delayed_replication_test/
http://cassci.datastax.com/view/trunk/job/trunk_dtest/346/testReport/junit/ttl_test/TestDistributedTTL/ttl_is_respected_on_repair_test/

After some debugging, I noticed a strange behaviour. It looks like some data disappear after a node restart, even if the row has no TTL set. Here a test example where I see the issue with latest trunk:

https://gist.github.com/aboudreault/94cb552750a186ca853d

Issue Links

• is related to: CASSANDRA-9669 If sstable flushes complete out of order, on restart we can fail to replay necessary commit log records