For each deployed container, a background process is started in /sebs/local/local.py to measure memory consumption by reading the cgroups memory.current file every few ms (can be configured with the --measure-interval flag).
When sebs stop is called, the processes are killed and for each container, the number of measurements, average and maximum memory consumption as well as a full profile are pretty printed to a json file.
mcopik
commented
1 year ago
Implemented changes discussed with Marcin.
For each deployed container, a background process is started in /sebs/local/local.py to measure memory consumption by reading the cgroups memory.current file every few ms (can be configured with the --measure-interval flag). When sebs stop is called, the processes are killed and for each container, the number of measurements, average and maximum memory consumption as well as a full profile are pretty printed to a json file.