one tikv has no leader and slow down br restore speed

[fubinzh]

Bug Report

Please answer these questions before submitting your issue. Thanks!

1. Minimal reproduce step (Required)

1. use br to restore a full backup of 30TB 500k table cluster to a 20 TiKV nodes new cluster.

/br  restore  full "-s" "gcs://aaa/bbb-500k-table-30tb-backup-with-stats-3" "-u" "http://tc-pd:2379" "--check-requirements=false" "--ddl-batch-size=128" "--tikv-max-restore-concurrency=128" "--granularity=coarse-grained" "--send-credentials-to-tikv=true" "--checksum=false"

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

all tikv should have restore tasks.

3. What did you see instead (Required)

One TiKV has no leader, and slow down the br restore speed.

4. What is your TiDB version? (Required)

[root@br-0 tmp]# /br -V
Release Version: v7.6.0-pr50518
Git Commit Hash: 4888198
Git Branch: heads/refs/tags/v7.6.0-pr50518
Go Version: go1.21.5
UTC Build Time: 2024-01-18 02:50:49
Race Enabled: false

[fubinzh]

/component br
/severity major

[3pointer]

This is because evict-slow-store triggeered by PD. we can fix this issue by remove this scheduler during restore.