Wrong memory usage when run program twice in one box

[HeRaNO]

Prepare the codes from #134 and run them continuously.

$ isolate --cg --cleanup
$ isolate --cg --init
/var/local/lib/isolate/0
$ isolate --cg -v -v -p --dir=box=/work --meta=metafile --run -- /usr/lib/jvm/java-11-openjdk-amd64/bin/java Main
Using control group box-0 under parent .
Started proxy_pid=... box_pid=... box_pid_inside_ns=...
Entering control group box-0
CG: Write tasks = 2
CG: Write tasks = 2
CG: Write tasks = 2
CG: Write cpuacct.usage = 0
Binding ...
...
OK (0.068 sec real, 0.072 sec wall)
$ cat metafile
...
max-rss:34144
...
cg-mem:10216
exitcode:0
$ isolate --cg -v -v -p --dir=box=/work --meta=metafile --run -- ./nul
Using control group box-0 under parent .
Started proxy_pid=... box_pid=... box_pid_inside_ns=...
Entering control group box-0
CG: Write tasks = 2
CG: Write tasks = 2
CG: Write tasks = 2
CG: Write cpuacct.usage = 0
Binding ...
OK (0.001 sec real, 0.025 sec wall)
$ cat metafile
...
max-rss:1152
...
cg-mem:10216
exitcode:0

It seems that the cg-mem has not been reset to 0. The isolate version is 1.10.1 with cgroups v1 enable.

UPD: Oops, we should set --cg-mem to make max_usage reset to 0. Perhaps we can reset it at the initializing?

[gollux]

This is expected behavior. If you run multiple commands in the same sandbox, the cgroup persists and it can retain cached data from the previous commands. I'm going to add a note to the man page, so that people are not surprised.

Generally, I recommend to use a fresh sandbox for each command.