[
https://issues.apache.org/jira/browse/CASSANDRA-10137?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14704833#comment-14704833
]
Benedict commented on CASSANDRA-10137:
--------------------------------------
You have used LWT by using "IF NOT EXISTS" with your insert. As a result, to be
certain your DELETE will apply, you must use IF EXISTS (or some other condition)
Any cell should only be updated by either LWT, or regular updates. If you mix
the two, you may see non-LWT updates disappear.
> Consistency problem
> -------------------
>
> Key: CASSANDRA-10137
> URL: https://issues.apache.org/jira/browse/CASSANDRA-10137
> Project: Cassandra
> Issue Type: Bug
> Reporter: Sergey
>
> I have 2 dc and 3 node:
> dc1: 2 node;
> dc2: 1 node;
> Exist keyspace
> KEYSPACE itm_dhcp_test WITH replication = {'class':
> 'NetworkTopologyStrategy', 'DC1': '2', 'DC2': '1'} AND durable_writes = true;
> and CF:
> TABLE itm_dhcp_test.lock (
> name text PRIMARY KEY,
> reason text,
> time timestamp,
> who text
> ) WITH bloom_filter_fp_chance = 0.01
> AND caching = '{"keys":"ALL", "rows_per_partition":"NONE"}'
> AND comment = ''
> AND compaction = {'min_threshold': '4', 'class':
> 'org.apache.cassandra.db.compaction.LeveledCompactionStrategy',
> 'max_threshold': '32'}
> AND compression = {'sstable_compression':
> 'org.apache.cassandra.io.compress.LZ4Compressor'}
> 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 = '99.0PERCENTILE';
> Periodically there is a problem with deleting records. For example execute
> query:
> INSERT INTO lock (name, reason, time, who) values ('unitTest4',
> 'CassandraClusterLockTest', dateof(now()), 'I') IF NOT EXISTS USING TTL 60
> SELECT * FROM lock WHERE name='unitTest4'
> DELETE FROM lock WHERE name='unitTest4'
> SELECT * FROM lock WHERE name='unitTest4'
> 20% - 30% of cases last SELECT returns not empty record. Most often when
> coordinator node1-dc2.
> In trace I see the message:
> | Parsing DELETE FROM lock WHERE name='unitTest4' | node1.dc2 |
> 45 | SharedPool-Worker-3
> | Preparing statement |
> node1.dc2 | 151 | SharedPool-Worker-3
> | Executing single-partition query on users |
> node1.dc2 | 588 | SharedPool-Worker-1
> | Acquiring sstable references |
> node1.dc2 | 601 | SharedPool-Worker-1
> | Merging memtable tombstones |
> node1.dc2 | 634 | SharedPool-Worker-1
> | Key cache hit for sstable 2 |
> node1.dc2| 668 | SharedPool-Worker-1
> | Seeking to partition beginning in data file |
> node1.dc2 | 674 | SharedPool-Worker-1
> | Skipped 0/1 non-slice-intersecting sstables, included 0 due to tombstones |
> node1.dc2 | 737 | SharedPool-Worker-1
> | Merging data from memtables and 1 sstables |
> node1.dc2 | 743 | SharedPool-Worker-1
> | Read 1 live and 0 tombstoned cells |
> node1.dc2 | 795 | SharedPool-Worker-1
> | Executing single-partition query on permissions |
> node1.dc2 | 1653 | SharedPool-Worker-1
> | Acquiring sstable references |
> node1.dc2 | 1662 | SharedPool-Worker-1
> | Merging memtable tombstones |
> node1.dc2 | 1690 | SharedPool-Worker-1
> | Key cache hit for sstable 5 |
> node1.dc2| 1737 | SharedPool-Worker-1
> | Seeking to partition indexed section in data file |
> node1.dc2 | 1742 | SharedPool-Worker-1
> | Skipped 0/1 non-slice-intersecting sstables, included 0 due to tombstones |
> node1.dc2 | 1797 | SharedPool-Worker-1
> | Merging data from memtables and 1 sstables |
> node1.dc2 | 1805 | SharedPool-Worker-1
> | Read 0 live and 0 tombstoned cells |
> node1.dc2 | 1819 | SharedPool-Worker-1
> | Executing single-partition query on users |
> node1.dc2 | 2798 | SharedPool-Worker-4
> | Acquiring sstable references |
> node1.dc2 | 2808 | SharedPool-Worker-4
> | Merging memtable tombstones |
> node1.dc2 | 2851 | SharedPool-Worker-4
> | Key cache hit for sstable 2 |
> node1.dc2| 2896 | SharedPool-Worker-4
> | Seeking to partition beginning in data file |
> node1.dc2 | 2903 | SharedPool-Worker-4
> | Skipped 0/1 non-slice-intersecting sstables, included 0 due to tombstones |
> node1.dc2 | 2948 | SharedPool-Worker-4
> | Merging data from memtables and 1 sstables |
> node1.dc2 | 2954 | SharedPool-Worker-4
> | Read 1 live and 0 tombstoned cells |
> node1.dc2 | 3004 | SharedPool-Worker-4
> | Executing single-partition query on permissions |
> node1.dc2 | 3807 | SharedPool-Worker-4
> | Acquiring sstable references |
> node1.dc2 | 3820 | SharedPool-Worker-4
> | Merging memtable tombstones |
> node1.dc2 | 3861 | SharedPool-Worker-4
> | Key cache hit for sstable 5 |
> node1.dc2| 3927 | SharedPool-Worker-4
> | Seeking to partition indexed section in data file |
> node1.dc2 | 3933 | SharedPool-Worker-4
> |
> Please help understand, why the record is not deleted?
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
