QPS periodically drops to the bottom when decrease info schema cache size

[lilinghai]

Bug Report

Please answer these questions before submitting your issue. Thanks!

1. Minimal reproduce step (Required)

2 million tables and dml on the 400k tables
the info cache usage is about 1GB, set the cache size 600MB after some time


The sql duration is long when qps drops to zero

[2025/01/07 06:40:19.476 +00:00] [WARN] [expensivequery.go:153] [expensive_query] [cost_time=480.664171214s] [cop_time=0.110959611s] [process_time=0.003311019s] [wait_time=0.105858558s] [request_count=3] [total_keys=262] [process_keys=242] [num_cop_tasks=3] [process_avg_time=0.001103673s] [process_p90_time=0.002575328s] [process_max_time=0.002575328s] [process_max_addr=[tc-tikv-5.tc](http://tc-tikv-5.tc/)-tikv-peer.forge-release2-tps-7693641-1-835.svc:20160] [wait_avg_time=0.035286186s] [wait_p90_time=0.105819697s] [wait_max_time=0.105819697s] [wait_max_addr=[tc-tikv-5.tc](http://tc-tikv-5.tc/)-tikv-peer.forge-release2-tps-7693641-1-835.svc:20160] [stats=sbtest2:454841630979784708] [conn=3311513750] [user=root] [database=test] [table_ids="[3259460]"] [index_names="[sbtest2:k_2]"] [txn_start_ts=455142680500895997] [mem_max="153182 Bytes (149.6 KB)"] [sql="select t1.k from sb100wnormal76488.sbtest2 as t1 left join sb100wnormal76488.sbtest2 as t2 on t1.k=t2.k where t1.k> 7652 and t1.k< 7751 order by t1.c limit 10"] [session_alias=] ["affected rows"=0]

2. What did you expect to see? (Required)

3. What did you see instead (Required)

4. What is your TiDB version? (Required)

Release Version: v9.0.0-alpha-33-gc44e9913a7
Edition: Community
Git Commit Hash: c44e991
Git Branch: HEAD
UTC Build Time: 2025-01-02 06:00:39
GoVersion: go1.23.4
Race Enabled: false
Check Table Before Drop: false
Store: unistore