This repo contains three functions. One support dynamic reload rules of prometheus, the other is used to support multi TiDB version.

Generate versioned monitoring data that is used by tidb-operator or tidb-ansible. It is a binary to load monitoring configurations from special component repo and it also contain platform monitoring from platform-monitoring directory.

Ansible deployment and operator deployment has different platform monitoring files. The structure of platform monitoring directory is here:

platform-monitoring/
     |-- ansbile
     |       |---- grafana
     |       |        | ---- machine.json
     |       |
     |       |---- rule
     |                | ---- machine.rules.yaml
     |
     |-- operator
             |---- grafana
             |        |
             |        |----- node.json
             |---- rule
             |        |
             |        |---- node.rules.yaml
             |---- datasource
                      |
                      |---- k8s-cluster.yaml
                      |
                      |---- tidb-cluster.yaml
           

The binary will automate load platform monitoring based on the above directory structure to different deployment directory.

go build -o pull-monitoring  cmd/monitoring.go

./pull-monitoring
Error: required flag(s) "config" not set
Usage:
  load [flags]

Flags:
      --auto-push                        auto generate new branch from master and push auto-generate files to the branch
      --config string                    the monitoring configuration file.
  -h, --help                             help for load
      --output-dir string                the base directory of the program (default ".")
      --tag string                       the tag of pull monitoring repo.

It is a simple binary to trigger a reload when Rules are updated. It watches dirs and call reload API that the rules has been changed.

It provide a UI to update rules(For ease of use, the UI is similar with UI of Prometheus).

The text editor is friendly to Yaml format. To quickly verify that the modification is successful, there is Rules UI which get rules from Prometheus (You can also verify it by Prometheus.).

make

There is binary in reload/build/{plateform}/reload, you can run it like this

./reload --watch-path=/tmp/prometheus-2.8.0.darwin-amd64/rules --prometheus-url=http://127.0.0.1:9090

Generate versioned monitoring data that is used by tidb-operator. This project pulls TiDB monitoring data from tidb-ansible and uses the git tag to understand the TiDB version.

All TiDB version monitoring information is automatically generated (by default it generates for TiDB version >= 2.1.8). The structure of monitor directory is like this

 monitor/
    |── v2.1.8
    |   ├── dashboards
    |   │   ├─ overview.json 
    |   │   ├─ binlog.json  
    |   │   |_ pd.json
    |   |   |_ tikv_pull.json
    |   |   |_ tidb.json 
    |   |   
    |   |── rules
    |   |   ├── tidb.rule.yml
    |   |   ├── tikv.rule.yml
    |   |   └── pd.rule.yml
    |   |—— Dockerfile     
    |   |__ init.sh
    |
    |── v3.0.0
    |   ├── dashboards
    |   │   |- overview.json 
    |   │   |- binlog.json  
    |   │   |- pd.json
    |   |   |- tidb.json 
    |   |   |- tikv_details.json
    |   |   |- tikv_sumary.json
    |   |   |_ tikv_trouble_shooting.json
    |   |   
    |   |── rules
    |   |   ├── tidb.rule.yml
    |   |   ├── tikv.rule.yml
    |   |   └── pd.rule.yml
    |   |—— Dockerfile     
    |   |__ init.sh
    |___ ...
        

make

There will be monitoring binary, you can run it like this

./monitoring --path=.

The program will replace some variables and the docker will receive 4 variables:

GF_PROVISIONING_PATH // grafana provisioning path
TIDB_CLUSTER_NAME // TiDB cluster name
TIDB_ENABLE_BINLOG // whether enable binlog
PROM_CONFIG_PATH // proemtheus rules config path

Reference PR: pingcap#227

1. Add dashbord files into the dashbords map in the file pkg/operator/dashboards.go, so that the platform-monitoring won't filter these files.

   // pkg/operator/dashboards.go
   dashboards = map[string]string{
   	  "binlog.json":                  "Test-Cluster-Binlog",
   	  "tidb.json":                    "Test-Cluster-TiDB",
   	  "overview.json":                "Test-Cluster-Overview",
     // ...
   }

   The key is the file name fetched from each github repository, and the value is the dashbord title in Grafana.

2. Update the init.sh of the image monitoring-initializer to provide new dashbord files for the Grafana.

   For example:

   # TiDB dashboard
   # copy dashboard files to the `$GF_PROVISIONING_PATH/dashboards` dir where Grafana loads dashboards.
   cp /tmp/tidb*.json $GF_PROVISIONING_PATH/dashboards  
   
   # replace some fields
   sed -i 's/Test-Cluster-TiDB/Cluster-TiDB/g' $GF_PROVISIONING_PATH/dashboards/tidb.json
   sed -i 's/Test-Cluster-TiDB/Cluster-TiDB/g' $GF_PROVISIONING_PATH/dashboards/tidb_runtime.json
   sed -i 's/Test-Cluster-TiDB/Cluster-TiDB/g' $GF_PROVISIONING_PATH/dashboards/tidb_resource_control.json

3. Build the image monitoring-initializer and test it locally.

   1. Build the binary pull-monitoring.

      make pull-monitoring

   2. Fetch dashbords files and generate the dir used to build a new image.

      ./pull-monitoring --config=monitoring.yaml --tag=${tag_in_target_repo} --token=${your_github_token}

      You can see a new dir named ${tag} created in the dir monitor-snapshot.

      $ ls ./monitor-snapshot/${tag}/operator
      dashboards  datasources  Dockerfile  init.sh  rules

   3. Build the image monitoring-initializer

   cd ./monitor-snapshot/master/operator &&  docker build -t ${image} .

   4. Follow the doc and use your image to deploy monitoring.

      Note that you need replace the image pingcap/tidb-monitor-initializer and version with your image in the TidbMonitor spec.