[ https://issues.apache.org/jira/browse/CASSANDRA-7066?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14681468#comment-14681468 ]
Stefania commented on CASSANDRA-7066: ------------------------------------- Hi Benedict, thank you for your review. I still need to self-review this and check CI results tomorrow, but if you want to quickly review in parallel to speed things up here is what I've done: * Added a new ABORT record that we use in {{TransactionTidier}} before removing unfinished leftovers of a 'live' transaction whose files are about to be deleted. We *still assume abort unless there is a COMMIT record* when getting temporary files or when cleaning up on startup. Is this what you intended? * Moved transaction log files to the same folder as the sstable files and renamed them to {{ma_txt_optype_id.log}} where {{ma}} is the latest sstable version. This ensures the txn logs sort at the end. When upgrading we will not recognize transactions for tables in an old format, which I assume is also a good thing? * Implemented unified listing of files using {{[newDirectoryStream|http://docs.oracle.com/javase/7/docs/api/java/nio/file/Files.html#newDirectoryStream(java.nio.file.Path)]}}, which allows us to check if the stream is atomic by checking if we have an instance of {{[SecureDirectoryStream|http://docs.oracle.com/javase/7/docs/api/java/nio/file/SecureDirectoryStream.html]}}, third paragraph in the documentation of {{newDirectoryStream}}. For the sake of clearness, once this ticket is committed, people currently running 3.0 will have to wipe the data folder, due to the compression metadata format changes and to the fact that old transaction log files will be ignored and will not get deleted automatically. > 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)