[ https://issues.apache.org/jira/browse/CASSANDRA-7066?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14578794#comment-14578794 ]
Tupshin Harper commented on CASSANDRA-7066: ------------------------------------------- Users don't care about SSTables, users care about their data. It's unclear what, if any, impact this would have on the availability/existence of data. So a few questions about failure conditions, all of which would apply to a single node cluster, and with commitlog durability set to batch, for simplicity of discussion. Could this result in any circumstances where: # a write was acknowledged to be written (consistency level met), but then no longer exists on disk through this sstable cleanup/deletion? # a datum was queryable (through memtable or sstable read), but then is either no longer on disk or queryable? # a datum was deleted (tombstone?) and then comes back? # similar questions to above when a snapshot/backup occurred prior to the sstable cleanup, and restoration from that backup was necessary. If the answer to all of those is "no", then I have a hard time imagining any objections, though would love additional input from others. If yes, then huge problem. :) Given the reference to "partial results" above, I'd also like some clarity on whether that has had any user-facing impact of data availability/queryability. > 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: compaction > Fix For: 3.x > > 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 that have been superceded); and 2) it's only used for a regular compaction - 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)