Following a conversation, last week, started about the implementation of the load balancing of HIVE nodes depending on the user's geolocation, I thought it would be nice to share with the HIVE developer community a Typescript service I made and use in some of my applications (Backend & Frontend).
After a few days to extract my service and transform it into an installable package (what I am not used to) I can finally give you the HIVE Nodes Checker.
Regularly checks availability and performance of HIVE blockchain RPC nodes asynchronously, stores/updates the result in an in-memory array, and sends the result through a subscription (RxJS).
In this mode, only one call to the condenser_api.get_config method is made on each node, The version executed on the node is retrieved and the response time is calculated by hrtime.
The message sent by RxJS is:
[{
"url": string
"nb_ok": number
"nb_error": number
"error"?: string
"last_check": number
"status": "unkown"|"online"|"degraded"|"error"
"duration"?: number
"average_time"?: number
"version"?: string
"deactivated"?: boolean
"test_result": []
}]
nb_ok and nb_error are incremented each time, allowing you to calculate an efficiency rate.
In this case the methods below are checked:
condenser_api.get_config
condenser_api.get_dynamic_global_properties
condenser_api.get_block
condenser_api.get_accounts
condenser_api.get_account_history
bridge.get_post
bridge.get_account_posts
The version executed on the node is retrieved, the response time is calculated by hrtime and an average answer time is calculated.
The message sent by RxJS is:
[{
"url": string
"nb_ok": number
"nb_error": number
"error"?: string
"last_check": number
"status": "unkown"|"online"|"degraded"|"error"
"duration"?: number
"average_time"?: number
"version"?: string
"deactivated"?: boolean
"test_result": [{
"name": string
"description": string
"method": string
"success": boolean
"duration": number
"error"?: string
"last_check": number
}]
}]
url: array of node url to checkoptions (optional):
full (boolean): if false perform only a get_config with response time (default: false)interval (seconds): delay in seconds between each execution (default: 900 seconds)timeout (seconds): timeout in seconds for node request (default: 3 seconds)Performing a full checking:
import { HiveNodesChecker } from '@mintrawa/hive-nodes-checker'
/** HIVE nodes url to check */
const nodes = [
"https://api.hive.blog",
"https://api.openhive.network",
"https://api.deathwing.me",
"https://rpc.ausbit.dev",
"https://rpc.ecency.com",
"https://hive-api.arcange.eu",
]
/** Options */
const options = {
full: true,
interval: 600 // 10 minutes
}
/** Init the checker */
const nodesChecker = new HiveNodesChecker(nodes, options)
/** Start the checker */
nodesChecker.start()
/** Subscribe results */
nodesChecker.message.subscribe({
next: async m => {
console.log('=====> NEW MESSAGE', m)
console.log()
},
error: error => {
console.log('=====> ERROR MESSAGE')
console.log(error)
console.log()
}
})
Github: https://github.com/Mintrawa/hive-nodes-checker
NPM: https://www.npmjs.com/package/@mintrawa/hive-nodes-checker
Tested in NodeJS v16.13.1 with a Raspberry PI3 1Gb
Tested in Angular v13.1.1
Tested in Chrome browser (HTML with the minified version)
Original photo by Lars Kienle on Unsplash
My HIVE witness servers
| CPU:Intel Xeon E3-1270v6 | CPU: Intel Xeon E3-1230v6 |
| 4 Cores/8 Threads 3.5 GHz/3.9 GHz | 4 Cores/8 Threads 3.8 GHz/4.2 GHz |
| RAM: 32GB DDR4 ECC 2133MHz | RAM: 32GB DDR4 ECC 2133MHz |
| HDD: 1 To SSD NVMe | HDD: 1 To SSD NVMe |
Vote for my HIVE witness: click here (via HiveSigner)