cgroup memory limit is not awared in some scenarios

[amyangfei]

What did you do?

• Deploy ticdc server with tiup
• Add MemoryMax=15G in systemd configuration.

$ cat /etc/systemd/system/cdc-8301.service
[Unit]
Description=cdc service
After=syslog.target network.target remote-fs.target nss-lookup.target

[Service]
LimitNOFILE=1000000
LimitSTACK=10485760
User=apple
ExecStart=/bin/bash -c '/home/apple/deploy/cdc-upstream-1/tidb-deploy/cdc-8301/scripts/run_cdc.sh'
Restart=always

RestartSec=15s
MemoryMax=15G

[Install]
WantedBy=multi-user.target

• Restart ticdc server

What did you expect to see?

The memory limit of cdc server should be 15G

Instead it uses host memory as total memory, and sets memory limit to 6.3G in sorter engine

[2023/03/20 12:15:40.097 +08:00] [INFO] [server.go:210] ["sorter engine memory limit"] [memory="6.3 GB"]

$ free -h
               total        used        free      shared  buff/cache   available
Mem:            58Gi       6.4Gi        47Gi        15Mi       5.2Gi        51Gi
Swap:             0B          0B          0B

What did you see instead?

TiCDC server doesn't find the correct memory limit, and uses total memory of host instead.

It is also a known bug in TiDB library: pingcap/tidb#38659

Versions of the cluster

TiCDC version (execute cdc version):

master@41b40aa