search

siemens / fluffi

FLUFFI (Fully Localized Utility For Fuzzing Instantaneously) - A distributed evolutionary binary fuzzer for pentesters
MIT License
99 stars 21 forks source link

Running no. of Local Managers are not consistent #256

Open vinamrabhatia opened 3 years ago

vinamrabhatia commented 3 years ago

We have observed in a number of instances that the running instances of Local Managers sometimes go usually high ( or they don't converge when we set all of them to 0 in Systems). This particular case happened when I made all the running instances as 0.

Screenshot_20210306_183206

The log message that we usually observed in the case is the following: unnamed_1)

On looking at the system logs, we observe Local Manager waiting for the Global Manager to assign a task to it. Besides, on running htop in the worker, we could actually see those Local Managers running as processes.

Furthermore, for the case when Local Manager didn't converge to 0 and stayed 1 after making running instances as 0, on adding a fuzzjob, the zombie Local Manager was actually picked by the new fuzz job and the system became consistent after that.

TomSie commented 3 years ago

I am not sure if that is your problem, but a known problem is that if there is no local manager for a fuzzjob, the displayed amount of instances for a system might be much higher than the actual number. This is due to the fact that the web gui just looks at what instances exist in the fuzzjob's database. If, however, there is no active Localmanager that checks if these instances are actually alive, the db will not reflect the reality.