Automatically troubleshooting SQL performance problems
#14 892 ouverte le 21 févr. 2020
Métriques du dépôt
- Stars
- (40 090 stars)
- Métriques de merge PR
- (Merge moyen 14j 4h) (346 PRs mergées en 30 j)
Description
Feature Request
Is your feature request related to a problem? Please describe:
Now that we have statement summary tables, we can query these tables by ourselves when there's performance degression. But the metrics in these tables are too many to figure out the problem quickly.
Describe the feature you'd like:
We can define diagnosis rules to find the problems by TiDB server itself. For example, when TiDB notices that the average latency of one kind of SQL increases sharply in the last 30 minutes, the server can compare each metric between the latest summary with any history summary. And it can figure out the problem, say, locking contention, execution plan changing. Once it figures out these problems, it can output it to a performance view, or a report.
Describe alternatives you've considered:
Teachability, Documentation, Adoption, Migration Strategy: