There are a lot of logs about "limit executor close takes a long time"

[crazycs520]

# rg "limit executor close takes a long time" tidb.log
12789:[2024/04/19 19:10:22.385 +08:00] [INFO] [executor.go:1424] ["limit executor close takes a long time"] [elapsed=2.107365ms]
12791:[2024/04/19 19:10:22.628 +08:00] [INFO] [executor.go:1424] ["limit executor close takes a long time"] [elapsed=1.822869ms]
12801:[2024/04/19 19:10:23.781 +08:00] [INFO] [executor.go:1424] ["limit executor close takes a long time"] [elapsed=4.897655ms]
...
13561:[2024/04/19 19:15:47.328 +08:00] [INFO] [executor.go:1424] ["limit executor close takes a long time"] [elapsed=1.374016ms]
...

# rg "limit executor close takes a long time" tidb.log | wc
    205    2665   25251

Look like the log is useless, can we remove it?

TiDB Version

nightly, 72e5460

[crazycs520]

@tiancaiamao PTAL

[time-and-fate]

This log has been useful in some internal emulation workloads. Maybe we can increase the threshold instead of removing it directly.

Related:
#50377
#40441 (comment)

[crazycs520]

Because this log contains so little information, there is no conn id, txn_ts, nor can we locate which query caused this.

If it's for performance optimization, might it be more useful to grab a profile with proof?

If it is use to locate a slow query, it is more appropriate to obtain more information, such as conn id, txn_ts, etc.?

[tiancaiamao]

Maybe we can increase the threshold instead of removing it directly.

+1

What's the workload you are running? @crazycs520