cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Stefania (JIRA)" <>
Subject [jira] [Commented] (CASSANDRA-7066) Simplify (and unify) cleanup of compaction leftovers
Date Tue, 09 Jun 2015 12:32:01 GMT


Stefania commented on CASSANDRA-7066:

Thank you for your review Benedict. I should be able to resume this in a couple of days.

Regarding SSTableDeletingTask, I'll see if I can remove it.

bq. Why does SSTableWriter.finish() take a TxnLog? For convenience/clarity? Perhaps we should
have a wrapped SSTableWriter that encapsulates this, so we don't confuse matters when we have
another encapsulation (Rewriter, etc)?

Pure convenience. I'll introduce a wrapper.

bq. I wonder if it might be more robust to only log the descriptors (except the opposing log
file, which would be a slight special case, but acceptable since it would be logged first,
so easily disambiguated). Just seems like we may be less likely to have some mistake that
leaves components lying around that we should have cleaned up.

The current strategy is to log any file that we might create, whether it exists or not, and
to delete files only if they exist. Logging a descriptor would therefore be equivalent, save
space in the log, and add clarity. We also need to know the relative path, because a transactions
folder may be in a location that is not a child of the data folder. I've seen it happen in
some unit tests. But I think we could just log the relative path and the descriptor base name
and that should be fine.

> Simplify (and unify) cleanup of compaction leftovers
> ----------------------------------------------------
>                 Key: CASSANDRA-7066
>                 URL:
>             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or conversely not cleanup files that have been superceded); and 2) it's only used for a regular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

View raw message