PD/DDL: Missing necessary metrics from etcd to identify the root cause of High 'Update Self Version Duration' when modifying column
- Dominant language
- Go
- Stars
- 40.5k
- Forks
- 6.2k
- PR merge metrics
- PR metrics pending
Description
## Bug Report
Please answer these questions before submitting your issue. Thanks!
### 1. Minimal reproduce step (Required)
Run continuous column modification DDLs on a TiDB cluster.
### 2. What did you expect to see? (Required)
The "Update Self Version Duration" metric in the DDL subsystem should remain low and stable.
### 3. What did you see instead (Required)
The "Update Self Version Duration" metric consistently peaks at around 9 minutes.

There is no clear cause for this delay based on etcd metrics.


Note: etcd’s disk WAL fsync rate does increase during this period, but it shouldn't be enough to cause such a dramatic delay. And the continues MVCC and other gRPC durations remain normal and stable.
### 4. What is your TiDB version? (Required)
v7.5.6
Contributor guide
Assessment
This issue has not been assessed yet.