Skip to content

Latest commit

 

History

History
283 lines (240 loc) · 8.15 KB

consul_kv.md

File metadata and controls

283 lines (240 loc) · 8.15 KB
Raw
title
consul_kv

Summary

For users that are using nginx-upsync-module and Consul KV as a service discovery, like the Weibo Mobile Team, this may be needed.

Thanks to @fatman-x guy, who developed this module, called consul_kv, and its worker process data flow is below: consul kv module data flow diagram

Configuration for discovery client

Configuration for Consul KV

Add following configuration in conf/config.yaml :

discovery:
  consul_kv:
    servers:
      - "http://127.0.0.1:8500"
      - "http://127.0.0.1:8600"
    prefix: "upstreams"
    skip_keys:                    # if you need to skip special keys
      - "upstreams/unused_api/"
    timeout:
      connect: 1000               # default 2000 ms
      read: 1000                  # default 2000 ms
      wait: 60                    # default 60 sec
    weight: 1                     # default 1
    fetch_interval: 5             # default 3 sec, only take effect for keepalive: false way
    keepalive: true               # default true, use the long pull way to query consul servers
    default_server:               # you can define default server when missing hit
      host: "127.0.0.1"
      port: 20999
      metadata:
        fail_timeout: 1           # default 1 ms
        weight: 1                 # default 1
        max_fails: 1              # default 1
    dump:                         # if you need, when registered nodes updated can dump into file
       path: "logs/consul_kv.dump"
       expire: 2592000      # unit sec, here is 30 day

And you can config it in short by default value:

discovery:
  consul_kv:
    servers:
      - "http://127.0.0.1:8500"

The keepalive has two optional values:

  • true, default and recommend value, use the long pull way to query consul servers
  • false, not recommend, it would use the short pull way to query consul servers, then you can set the fetch_interval for fetch interval

Dump Data

When we need reload apisix online, as the consul_kv module maybe loads data from CONSUL slower than load routes from ETCD, and would get the log at the moment before load successfully from consul:

 http_access_phase(): failed to set upstream: no valid upstream node

So, we import the dump function for consul_kv module. When reload, would load the dump file before from consul; when the registered nodes in consul been updated, would dump the upstream nodes into file automatically.

The dump has three optional values now:

  • path, the dump file save path
    • support relative path, eg: logs/consul_kv.dump
    • support absolute path, eg: /tmp/consul_kv.bin
    • make sure the dump file's parent path exist
    • make sure the apisix has the dump file's read-write access permission,eg: chown www:root conf/upstream.d/
  • load_on_init, default value is true
    • if true, just try to load the data from the dump file before loading data from consul when starting, does not care the dump file exists or not
    • if false, ignore loading data from the dump file
    • Whether true or false, we don't need to prepare a dump file for apisix at anytime
  • expire, unit sec, avoiding load expired dump data when load
    • default 0, it is unexpired forever
    • recommend 2592000, which is 30 days(equals 3600 * 24 * 30)

Register Http API Services

Service register Key&Value template:

Key:    {Prefix}/{Service_Name}/{IP}:{Port}
Value: {"weight": <Num>, "max_fails": <Num>, "fail_timeout": <Num>}

The register consul key use upstreams as prefix by default. The http api service name called webpages for example, and you can also use webpages/oneteam/hello as service name. The api instance of node's ip and port make up new key: <IP>:<Port>.

Now, register nodes into consul:

curl \
    -X PUT \
    -d ' {"weight": 1, "max_fails": 2, "fail_timeout": 1}' \
    http://127.0.0.1:8500/v1/kv/upstreams/webpages/172.19.5.12:8000

curl \
    -X PUT \
    -d ' {"weight": 1, "max_fails": 2, "fail_timeout": 1}' \
    http://127.0.0.1:8500/v1/kv/upstreams/webpages/172.19.5.13:8000

In some case, same keys exist in different consul servers. To avoid confusion, use the full consul key url path as service name in practice.

Upstream setting

Here is an example of routing a request with a URL of "/*" to a service which named "http://127.0.0.1:8500/v1/kv/upstreams/webpages/" and use consul_kv discovery client in the registry :

$ curl http://127.0.0.1:9180/apisix/admin/routes/1 -H 'X-API-KEY: edd1c9f034335f136f87ad84b625c8f1' -X PUT -i -d '
{
    "uri": "/*",
    "upstream": {
        "service_name": "http://127.0.0.1:8500/v1/kv/upstreams/webpages/",
        "type": "roundrobin",
        "discovery_type": "consul_kv"
    }
}'

The format response as below:

{
  "node": {
    "value": {
      "priority": 0,
      "update_time": 1612755230,
      "upstream": {
        "discovery_type": "consul_kv",
        "service_name": "http://127.0.0.1:8500/v1/kv/upstreams/webpages/",
        "hash_on": "vars",
        "type": "roundrobin",
        "pass_host": "pass"
      },
      "id": "1",
      "uri": "/*",
      "create_time": 1612755230,
      "status": 1
    },
    "key": "/apisix/routes/1"
  }
}

You could find more usage in the apisix/t/discovery/consul_kv.t file.

Debugging API

It also offers control api for debugging.

Memory Dump API

GET /v1/discovery/consul_kv/dump

For example:

# curl http://127.0.0.1:9090/v1/discovery/consul_kv/dump | jq
{
  "config": {
    "fetch_interval": 3,
    "timeout": {
      "wait": 60,
      "connect": 6000,
      "read": 6000
    },
    "prefix": "upstreams",
    "weight": 1,
    "servers": [
      "http://172.19.5.30:8500",
      "http://172.19.5.31:8500"
    ],
    "keepalive": true,
    "default_service": {
      "host": "172.19.5.11",
      "port": 8899,
      "metadata": {
        "fail_timeout": 1,
        "weight": 1,
        "max_fails": 1
      }
    },
    "skip_keys": [
      "upstreams/myapi/gateway/apisix/"
    ]
  },
  "services": {
    "http://172.19.5.31:8500/v1/kv/upstreams/webpages/": [
      {
        "host": "127.0.0.1",
        "port": 30513,
        "weight": 1
      },
      {
        "host": "127.0.0.1",
        "port": 30514,
        "weight": 1
      }
    ],
    "http://172.19.5.30:8500/v1/kv/upstreams/1614480/grpc/": [
      {
        "host": "172.19.5.51",
        "port": 50051,
        "weight": 1
      }
    ],
    "http://172.19.5.30:8500/v1/kv/upstreams/webpages/": [
      {
        "host": "127.0.0.1",
        "port": 30511,
        "weight": 1
      },
      {
        "host": "127.0.0.1",
        "port": 30512,
        "weight": 1
      }
    ]
  }
}

Show Dump File API

It offers another control api for dump file view now. Maybe would add more api for debugging in future.

GET /v1/discovery/consul_kv/show_dump_file

For example:

curl http://127.0.0.1:9090/v1/discovery/consul_kv/show_dump_file | jq
{
  "services": {
    "http://172.19.5.31:8500/v1/kv/upstreams/1614480/webpages/": [
      {
        "host": "172.19.5.12",
        "port": 8000,
        "weight": 120
      },
      {
        "host": "172.19.5.13",
        "port": 8000,
        "weight": 120
      }
    ]
  },
  "expire": 0,
  "last_update": 1615877468
}