loader failed with "restoring count greater than total count for table"

[D3Hunter]

What did you do?

only one worker in this scenario

1. start task A in all mode
2. start relay

• task A finished load and get into sync, so all dumped files are removed, but before sync starts it wait at unitTransWaitCondition(inside SubTask.run()) for relay is enabled, so got no chance to save a sync checkpoint
• then EnableRelay starts to resetAllSubTasks, a new load unit is created for it's the last unit that is not fresh and begin running
• load found no dumped files but got one file from checkpoint(already finished), then reported an error(in CalcProgress)

not that easy to repeat this bug. i write a script to start 4 tasks and start relay immediately may repeat it sometimes.

What did you expect to see?

no error

What did you see instead?

error

Versions of the cluster

DM version (run dmctl -V or dm-worker -V or dm-master -V):

# bin/dm-master  -V
Release Version: v2.0.0-beta.2-555-gaf5f7e2-dev
Git Commit Hash: af5f7e2d8f63deda79812140b3c414474be7efd3
Git Branch: relay-notify
UTC Build Time: 2021-10-18 10:59:58
Go Version: go version go1.17.2 linux/amd64

current status of DM cluster (execute query-status in dmctl)

(paste current status of DM cluster here)