Mutual influence when different CPU are fullload with experimental destroy

[86542822]

The server has N CPUs
Mount one of the CPUs fullload
./blade create cpu fullload --numcpu 1
{"code":200,"success":true,"result":"51d50db1faf4fb80"}

Then mount one of the CPUs fullload
./blade create cpu fullload --numcpu 1
{"code":200,"success":true,"result":"96a4b01617f5f2f7"}

Destroy one of the mount experiments
./blade destroy 51d50db1faf4fb80

At this time, both CPU mount experiments will be destroyed.

[Super-long]

I get it!

I figured out the problem after I followed the dlv to debug the entire life cycle of a destroy. This bug appears in chaosblade-exec-os instead of chaosblade, in fact the parameter that blade happens first to chaos_os is completely correct, but chaos_os screwed everything up,

The call stack of the problem is like this main.main -> cpu.(*cpuExecutor).Exec -> cpu.(*cpuExecutor).stop -> exec.Destroy. The uid setting in main.main is already wrong, But the actual problem is in exec.Destroy, Destroy does not consider uid at all when deleting exp, and directly deletes all processes with "chaos_os" and "cpu fullload" strings in Precossname.

I am solving this problem.

[Super-long]

My idea is to divide into two steps, the first step is to correctly parse the uid in main.main, and the second is to add a function to delete the specified uid process in chaosblade-exec-os/exec/exec.go. what do you think of mr xcaspar? @xcaspar

[Super-long]

hello everyone, I put the solution in chaosblade-io/chaosblade-spec-go#44

[Super-long]

This is a new solution. chaosblade-io/chaosblade-exec-os#114 @tiny-x @xcaspar

[Icesource]

@Super-long Could you please help me review the PR chaosblade-io/chaosblade-exec-os#118 ?