[ https://issues.apache.org/jira/browse/CASSANDRA-7066?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14648509#comment-14648509 ]
Benedict commented on CASSANDRA-7066: ------------------------------------- [~rcoli]: while this patch does not remove the ancestor maintenance, it does make it completely irrelevant, since it is no longer used for any behaviour. However your concern highlights a much scarier prospect: the replay positions stored within the sstables could be at an arbitrary position, potentially in the future of the node's commit log. So they could break commit log replay during startup. In fact they probably have a high chance of doing so, although CASSANDRA-9669 may mitigate that. We should actually remove this data from {{CompactionMetadata}}, and mention this in the NEWS.txt as well. bq. Is it possible to have the refresh command and possibly node startup fail and/or throw an error if this step doesn't happen? There's unfortunately no good way for the node to know that it should not replay these files, since the whole point of their existence is that they should be replayed. The changes we've just discussed will make us _somewhat_ robust to any problems that might arise from operators doing this, and the likelihood of copying files covered by such a file is fairly low anyway. [~stefania]: perhaps we should offer another command line tool that just performs the cleanup (i.e. either rolls back or finishes commit, whichever is necessary) that an operator can run to clear the way, and we can make this explicit in NEWS.txt > Simplify (and unify) cleanup of compaction leftovers > ---------------------------------------------------- > > Key: CASSANDRA-7066 > URL: https://issues.apache.org/jira/browse/CASSANDRA-7066 > Project: Cassandra > Issue Type: Improvement > Components: Core > Reporter: Benedict > Assignee: Stefania > Priority: Minor > Labels: benedict-to-commit, compaction > Fix For: 3.0 alpha 1 > > Attachments: 7066.txt > > > Currently we manage a list of in-progress compactions in a system table, > which we use to cleanup incomplete compactions when we're done. The problem > with this is that 1) it's a bit clunky (and leaves us in positions where we > can unnecessarily cleanup completed files, or conversely not cleanup files > that have been superceded); and 2) it's only used for a regular compaction - > no other compaction types are guarded in the same way, so can result in > duplication if we fail before deleting the replacements. > I'd like to see each sstable store in its metadata its direct ancestors, and > on startup we simply delete any sstables that occur in the union of all > ancestor sets. This way as soon as we finish writing we're capable of > cleaning up any leftovers, so we never get duplication. It's also much easier > to reason about. -- This message was sent by Atlassian JIRA (v6.3.4#6332)