[REVIEW NOTIFICATION]

This pull request has been approved by:

• mjonss
• xhebox

To complete the pull request process, please ask the reviewers in the list to review by filling /cc @reviewer in the comment.
After your PR has acquired the required number of LGTMs, you can assign this pull request to the committer in the list by filling /assign @committer in the comment to help you merge this pull request.

The full list of commands accepted by this bot can be found here.

The grammar should be discussed, as the DROP STATS t1, t2 GLOBAL and DROP STATS t PARTITION xxx is not consistent now. I'm not sure whether it's expected, or I shouldn't allow dropping multiple tables with GLOBAL at the end?

My short answer: Only allow DROP STATS t[,t2[...]]. No need to handle GLOBAL (always table/global level) or PARTITION (since we want to always have table/global stats and all partitions)

I have a patch for always enabling GLOBAL stats for partitioned tables here, which I also see as a part of this discussion :)

Due to the plans of removing static partition prune mode, I would prefer that we also deprecating the support of partition specific configuration of statistics, and not adding support of dropping stats for a specific partition.

Unless there are plans to change the statistic usage for partitioned tables under dynamic prune mode?

I propose that we only support ANALYZE table (global) with non-default configuration (like buckets, samples etc.) and only saves the configuration on table/global level.

My reasoning is that the partition level statistics is aggregated to table/global level and should not differ in configuration which may result in issues when calculating the table/global statistic.

Meaning the following syntax should be supported:
ANALYZE TABLE <list of tables> ... No change.
ANALYZE TABLE <partitioned table> PARTITION <list of partitions> [INDEX <list of indexes>] No additional configuration allowed or used, but still allowed to update the statistics from a subset of partitions.
ALTER TABLE <partitioned table> ANALYZE PARTITION <list of partitions> [INDEX <list of indexes> No additional configuration allowed or used, but still allowed to update the statistics from a subset of partitions.
DROP STATS <list of tables> (no mentioning of partitions allowed or supported)

And this would be deprecated:
ANALYZE TABLE <partitioned table> PARTITION <list of partitions> AllColumnsOrPredicateColumnsOpt AnalyzeOptionListOpt
ALTER TABLE <partitioned table> ANALYZE PARTITION <list of partitions> AllColumnsOrPredicateColumnsOpt AnalyzeOptionListOpt
ANALYZE INCREMENTAL TABLE <partitioned table> PARTITION...
ANALYZE TABLE <partitioned table> PARTITION <list of partitions> COLUMNS...

My short answer: Only allow DROP STATS t[,t2[...]]. No need to handle GLOBAL (always table/global level) or PARTITION (since we want to always have table/global stats and all partitions)

@mjonss Good answer! IIUC, DROP STATS ... GLOBAL and DROP STATS ... PARTITION ... will be deprecated? In the latest update of this patch, I add warnings for these two syntax, and don't touch the function of them (before they are useless). In conclusion, this PR becomes the following change:

1. Deprecate the DROP STATS ... GLOBAL and DROP STATS ... PARTITION ... syntax
2. Allow dropping stats for multiple tables through DROP STATS a,b,c,d

PTAL

@YangKeao rebase please, let us merge it.

/merge

This pull request has been accepted and is ready to merge.

@YangKeao Rebase please. Also, don't forget pingcap/doc changes.

I have updated the documents in pingcap/docs-cn#11986 and pingcap/docs#11287. Could you please tech-review these documents? @mjonss @xhebox . I only update the syntax (following the DROP TABLE ... style, didn't add more examples about dropping multiple tables). The DROP STATS ... PARTITION ... and DROP STATS ... GLOBAL are not updated there, because they are deprecated (though, have never been on the documents).

TiDB MergeCI notify

🔴 Bad News! New failing [1] after this pr merged.
These new failed integration tests seem to be caused by the current PR, please try to fix these new failed integration tests, thanks!