Artifacts accompanying HolisticRCA, a framework for failure root cause analysis in cloud-native systems from a holistic perspective.
cd ./code
pip install -r requirements.txt
Note: we have split the zip file for uploading. The following commands need to be input for the working folder:
zip model_split.zip -s=0 --out model.zip
unzip model.zip
Then the temp data also needs to be downloaded from https://github.com/baiyanquan/HolisticRCATempData.
Finally the model
and temp_data
folders need be placed following the structure of the working folder:
.
├── README.md
├── code
│ ├── data_filter preprocess data
│ │ ├── CCF_AIOps_challenge_2022 preprocess dataset A
│ │ ├── ICASSP_AIOps_challenge_2022 preprocess dataset B
│ │ └── Eadro_TT_and_SN preprocess dataset C
│ ├── HolisticRCA the main model of the work
│ │ ├── ablation models for ablation study
│ │ ├── base base classes for model construction
│ │ ├── config configuration of file paths
│ │ ├── data_loader load dataset
│ │ ├── dataset base class for dataset reader
│ │ ├── explain mask learning component (for resource entity localization and fault-related observability data localization)
│ │ ├── model main components except mask learning
│ │ ├── trainer perform model training
│ │ └── util RERG class and data transformation
│ ├── shared_util some basic util functions
│ ├── experiments_a.sh quick experiments for dataset A
│ ├── experiments_b.sh quick experiments for dataset B
│ ├── experiments_c.sh quick experiments for dataset C
│ └── requirements.txt
├── model saved model data for reproduction
└── temp_data saved temp data for reproduction
requirements.txt
.model.zip
and temp_data.zip
.The saved model files are placed in model.zip
. Following the files experiments_a.sh
, experiments_b.sh
, or experiments_c.sh
and comment out rca_data_trainer.train()
in the corresponding training files. It will output the evaluation results (note that some file paths need to be changed).
cd ./code
bash experiments_a.sh
bash experiments_b.sh
bash experiments_c.sh
Since the raw data is too big, we list their links here, help for downloading: