[
https://issues.apache.org/jira/browse/CASSANDRA-13992?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16252848#comment-16252848
]
Kurt Greaves commented on CASSANDRA-13992:
------------------------------------------
I don't see why special casing LWT is necessary. It couples LWT,
{{ModificationStatement}} and {{BatchStatement}} with {{ExecuteMessage}}, which
seems a bit messy. It also introduces a weird edge case into the protocol -
i.e,
1. you can't skip metadata for LWT
2. metadata will never be changed for LWT.
2 is fine as it's somewhat the goal of the ticket, but when we already have a
mechanism to meet this requirement it seems silly to wrap it in another. 1 just
adds complexity to the protocol, which imo, is complex enough. It's not even
clear in the protocol that SKIP_METADATA will be ignored if the metadata *did*
change (we should update this in the patch). I think this can be better solved
simply by providing a consistent MD5 for LWT's as my patch already does
[here|https://github.com/apache/cassandra/compare/trunk...kgreav:13992]. This
removes the need to introduce new imports, and special casing to
{{ExecuteMessage}} while achieving the same behaviour, and I think logically is
easier to explain.
AFAICT [~ifesdjeen]'s patch only really works because the driver is already set
up for it to work. The _java_ driver already checks that {{new_metadata_id !=
null}} which is logical, but in no way required. It seems to me we're
introducing a metadata_id that could potentially be null where a driver might
expect it to have a value. I prefer my patch, or better yet, we create an
{{EMPTY_DIGEST}} that's an actual MD5 to use so we don't have to rely on
{{compute}} every time we want to use an {{EMPTY_DIGEST}}
> Don't send new_metadata_id for conditional updates
> --------------------------------------------------
>
> Key: CASSANDRA-13992
> URL: https://issues.apache.org/jira/browse/CASSANDRA-13992
> Project: Cassandra
> Issue Type: Bug
> Reporter: Olivier Michallat
> Assignee: Kurt Greaves
> Priority: Minor
>
> This is a follow-up to CASSANDRA-10786.
> Given the table
> {code}
> CREATE TABLE foo (k int PRIMARY KEY)
> {code}
> And the prepared statement
> {code}
> INSERT INTO foo (k) VALUES (?) IF NOT EXISTS
> {code}
> The result set metadata changes depending on the outcome of the update:
> * if the row didn't exist, there is only a single column \[applied] = true
> * if it did, the result contains \[applied] = false, plus the current value
> of column k.
> The way this was handled so far is that the PREPARED response contains no
> result set metadata, and therefore all EXECUTE messages have SKIP_METADATA =
> false, and the responses always include the full (and correct) metadata.
> CASSANDRA-10786 still sends the PREPARED response with no metadata, *but the
> response to EXECUTE now contains a {{new_metadata_id}}*. The driver thinks it
> is because of a schema change, and updates its local copy of the prepared
> statement's result metadata.
> The next EXECUTE is sent with SKIP_METADATA = true, but the server appears to
> ignore that, and still sends the metadata in the response. So each response
> includes the correct metadata, the driver uses it, and there is no visible
> issue for client code.
> The only drawback is that the driver updates its local copy of the metadata
> unnecessarily, every time. We can work around that by only updating if we had
> metadata before, at the cost of an extra volatile read. But I think the best
> thing to do would be to never send a {{new_metadata_id}} in for a conditional
> update.
--
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
