[ https://issues.apache.org/jira/browse/CASSANDRA-13918?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16186459#comment-16186459 ]
Blake Eggleston commented on CASSANDRA-13918: --------------------------------------------- utests look good on all branches, dtests look good on 3.0 & 3.11. trunk dtests aborted, but looks like it was due to an existing issue. > Header only commit logs should be filtered before recovery > ---------------------------------------------------------- > > Key: CASSANDRA-13918 > URL: https://issues.apache.org/jira/browse/CASSANDRA-13918 > Project: Cassandra > Issue Type: Bug > Reporter: Blake Eggleston > Assignee: Blake Eggleston > Fix For: 3.0.15, 4.0, 3.11.x > > > Commit log recovery will tolerate commit log truncation in the most recent > log file found on disk, but will abort startup if problems are detected in > others. > Since we allocate commit log segments before they're used though, it's > possible to get into a state where the last commit log file actually written > to is not the same file that was most recently allocated, preventing startup > for what should otherwise be allowable incomplete final segments. > Excluding header only files on recovery should prevent this from happening. -- This message was sent by Atlassian JIRA (v6.4.14#64029) --------------------------------------------------------------------- To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org For additional commands, e-mail: commits-h...@cassandra.apache.org