[ https://issues.apache.org/jira/browse/CASSANDRA-14521?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16512434#comment-16512434 ]
Julien commented on CASSANDRA-14521: ------------------------------------ Priority decreased as I have a workaround. My application is able to know that the DELETE statement will be followed by an INSERT, and that it is therefore useless as long as the INSERT has all necessary columns. > With server-generated timestamps, INSERT after DELETE may not be applied > ------------------------------------------------------------------------ > > Key: CASSANDRA-14521 > URL: https://issues.apache.org/jira/browse/CASSANDRA-14521 > Project: Cassandra > Issue Type: Bug > Reporter: Julien > Priority: Minor > > We use server-generated timestamps for all requests because of > CASSANDRA-14304. > The scenario is basically the following: > {code} > INSERT INTO mytable(id) VALUES ('1'); > DELETE FROM mytable WHERE id='1'; > INSERT INTO mytable(id) VALUES ('1'); > SELECT * FROM mytable WHERE id='1'; > {code} > SELECT _sometimes_ does not return anything when the java driver has > {{CassandraClientConnector.with(ServerSideTimestampGenerator.INSTANCE);}} and > the Cassandra cluster has 3 nodes and replication-factor:3. > This scenario actually works as expected with CQL because I don't know how to > force the usage of server-generated timestamps with CQL. Is it possible? > It also works correctly with a single Cassandra node. -- This message was sent by Atlassian JIRA (v7.6.3#76005) --------------------------------------------------------------------- To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org For additional commands, e-mail: commits-h...@cassandra.apache.org