Skip to content

dragonly/tiflash_mpp_profiler

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

61 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Usage

# install from pypi
pip3 install flashprof
# collect tiflash logs from tiup cluster to current directory, which will also be parsed to json
flashprof collect --cluster $CLUSTER_NAME
# render all cluster runtime info, currently supports task DAG
flashprof render
# help
flashprof -h
flashprof <subcommand> -h

DAG

Currently only task runtime info is visualized as DAG. Tasks in a single query may span across several tiflash instances, and those with status != FINISHED or error_message != "" will be labelled with red border.

data layout

The collected/generated artifacts have the following layout

flashprof
└── cluster
    ├── cluster1_name
    │   ├── log (collected from tiflash log dir)
    │   │   ├── ip1.tiflash.log
    │   │   └── ip2.tiflash.log
    │   └── task_dag (parsed and combined task dag)
    │       ├── json
    │       │   ├── ip1.tiflash.log.task_dag.json
    │       │   ├── ip2.tiflash.log.task_dag.json
    │       │   └── cluster.task_dag.json
    │       ├── png (rendered png files)
    │       └── svg (rendered svg files)
    └── cluster2_name
...

Development

# install a local dev version of python package, then we can call flashprof
# rerun this when code is changed
# it internally creates a symbolic link to the current source code
pip3 install -e .
# remove if you want
pip3 uninstall flashprof

Packaging

TL;DR

pip3 install build
python3 -m build
twine check dist/*
twine upload dist/*
# upload to test.pypi.org for package publish related test
# twine upload --repository testpypi dist/*

Instructions

Please refer to https://packaging.python.org/guides/distributing-packages-using-setuptools/ for detaild instructions.

Internals

collect command collects tiflash logs according to the tiup configurations for the specified --cluster $CLUSTER_NAME, and logs are named $IP.tiflash.log in flashprof/cluster/$CLUSTER_NAME/log.

parse command parses all the tiflash logs collected above to the json format, which only contains task DAGs for now. The json files a then merged into a cluster.json in flashprof/cluster/$CLUSTER_NAME/task_dag/json.

render command renders cluster.json into dag graphs per query_tso in flashprof/cluster/$CLUSTER_NAME/$FORMAT.

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages