n8n/packages/cli/commands/webhook.ts

224 lines
6.6 KiB
TypeScript
Raw Normal View History

:sparkles: Separate webhooks from core (#1408) * Unify execution ID across executions * Fix indentation and improved comments * WIP: saving data after each node execution * Added on/off to save data after each step, saving initial data and retries working * Fixing lint issues * Fixing more lint issues * :sparkles: Add bull to execute workflows * :shirt: Fix lint issue * :zap: Add graceful shutdown to worker * :zap: Add loading staticData to worker * :shirt: Fix lint issue * :zap: Fix import * Changed tables metadata to add nullable to stoppedAt * Reload database on migration run * Fixed reloading database schema for sqlite by reconnecting and fixing postgres migration * Added checks to Redis and exiting process if connection is unavailable * Fixing error with new installations * Fix issue with data not being sent back to browser on manual executions with defined destination * Merging bull and unify execution id branch fixes * Main process will now get execution success from database instead of redis * Omit execution duration if execution did not stop * Fix issue with execution list displaying inconsistant information information while a workflow is running * Remove unused hooks to clarify for developers that these wont run in queue mode * Added active pooling to help recover from Redis crashes * Lint issues * Changing default polling interval to 60 seconds * Removed unnecessary attributes from bull job * Added webhooks service and setting to disable webhooks from main process * Fixed executions list when running with queues. Now we get the list of actively running workflows from bull. * Add option to disable deregistration of webhooks on shutdown * Rename WEBHOOK_TUNNEL_URL to WEBHOOK_URL keeping backwards compat. * Added auto refresh to executions list * Improvements to workflow stop process when running with queues * Refactor queue system to use a singleton and avoid code duplication * Improve comments and remove unnecessary commits * Remove console.log from vue file * Blocking webhook process to run without queues * Handling execution stop graciously when possible * Removing initialization of all workflows from webhook process * Refactoring code to remove code duplication for job stop * Improved execution list to be more fluid and less intrusive * Fixing workflow name for current executions when auto updating * :zap: Right align autorefresh checkbox Co-authored-by: Jan Oberhauser <jan.oberhauser@gmail.com>
2021-02-09 14:32:40 -08:00
import {
UserSettings,
} from 'n8n-core';
import { Command, flags } from '@oclif/command';
import * as Redis from 'ioredis';
import * as config from '../config';
import {
ActiveExecutions,
ActiveWorkflowRunner,
CredentialsOverwrites,
CredentialTypes,
Db,
ExternalHooks,
GenericHelpers,
LoadNodesAndCredentials,
NodeTypes,
TestWebhooks,
WebhookServer,
2021-03-25 03:23:54 -07:00
} from '../src';
:sparkles: Separate webhooks from core (#1408) * Unify execution ID across executions * Fix indentation and improved comments * WIP: saving data after each node execution * Added on/off to save data after each step, saving initial data and retries working * Fixing lint issues * Fixing more lint issues * :sparkles: Add bull to execute workflows * :shirt: Fix lint issue * :zap: Add graceful shutdown to worker * :zap: Add loading staticData to worker * :shirt: Fix lint issue * :zap: Fix import * Changed tables metadata to add nullable to stoppedAt * Reload database on migration run * Fixed reloading database schema for sqlite by reconnecting and fixing postgres migration * Added checks to Redis and exiting process if connection is unavailable * Fixing error with new installations * Fix issue with data not being sent back to browser on manual executions with defined destination * Merging bull and unify execution id branch fixes * Main process will now get execution success from database instead of redis * Omit execution duration if execution did not stop * Fix issue with execution list displaying inconsistant information information while a workflow is running * Remove unused hooks to clarify for developers that these wont run in queue mode * Added active pooling to help recover from Redis crashes * Lint issues * Changing default polling interval to 60 seconds * Removed unnecessary attributes from bull job * Added webhooks service and setting to disable webhooks from main process * Fixed executions list when running with queues. Now we get the list of actively running workflows from bull. * Add option to disable deregistration of webhooks on shutdown * Rename WEBHOOK_TUNNEL_URL to WEBHOOK_URL keeping backwards compat. * Added auto refresh to executions list * Improvements to workflow stop process when running with queues * Refactor queue system to use a singleton and avoid code duplication * Improve comments and remove unnecessary commits * Remove console.log from vue file * Blocking webhook process to run without queues * Handling execution stop graciously when possible * Removing initialization of all workflows from webhook process * Refactoring code to remove code duplication for job stop * Improved execution list to be more fluid and less intrusive * Fixing workflow name for current executions when auto updating * :zap: Right align autorefresh checkbox Co-authored-by: Jan Oberhauser <jan.oberhauser@gmail.com>
2021-02-09 14:32:40 -08:00
import { IDataObject } from 'n8n-workflow';
let activeWorkflowRunner: ActiveWorkflowRunner.ActiveWorkflowRunner | undefined;
let processExistCode = 0;
export class Webhook extends Command {
static description = 'Starts n8n webhook process. Intercepts only production URLs.';
static examples = [
`$ n8n webhook`,
];
static flags = {
help: flags.help({ char: 'h' }),
};
/**
* Stops the n8n in a graceful way.
* Make for example sure that all the webhooks from third party services
* get removed.
*/
static async stopProcess() {
console.log(`\nStopping n8n...`);
try {
const externalHooks = ExternalHooks();
await externalHooks.run('n8n.stop', []);
setTimeout(() => {
// In case that something goes wrong with shutdown we
// kill after max. 30 seconds no matter what
process.exit(processExistCode);
}, 30000);
const removePromises = [];
if (activeWorkflowRunner !== undefined) {
removePromises.push(activeWorkflowRunner.removeAll());
}
// Remove all test webhooks
const testWebhooks = TestWebhooks.getInstance();
removePromises.push(testWebhooks.removeAll());
await Promise.all(removePromises);
// Wait for active workflow executions to finish
const activeExecutionsInstance = ActiveExecutions.getInstance();
let executingWorkflows = activeExecutionsInstance.getActiveExecutions();
let count = 0;
while (executingWorkflows.length !== 0) {
if (count++ % 4 === 0) {
console.log(`Waiting for ${executingWorkflows.length} active executions to finish...`);
}
await new Promise((resolve) => {
setTimeout(resolve, 500);
});
executingWorkflows = activeExecutionsInstance.getActiveExecutions();
}
} catch (error) {
console.error('There was an error shutting down n8n.', error);
}
process.exit(processExistCode);
}
async run() {
// Make sure that n8n shuts down gracefully if possible
process.on('SIGTERM', Webhook.stopProcess);
process.on('SIGINT', Webhook.stopProcess);
const { flags } = this.parse(Webhook);
// Wrap that the process does not close but we can still use async
await (async () => {
if (config.get('executions.mode') !== 'queue') {
2021-03-25 03:23:54 -07:00
/**
:sparkles: Separate webhooks from core (#1408) * Unify execution ID across executions * Fix indentation and improved comments * WIP: saving data after each node execution * Added on/off to save data after each step, saving initial data and retries working * Fixing lint issues * Fixing more lint issues * :sparkles: Add bull to execute workflows * :shirt: Fix lint issue * :zap: Add graceful shutdown to worker * :zap: Add loading staticData to worker * :shirt: Fix lint issue * :zap: Fix import * Changed tables metadata to add nullable to stoppedAt * Reload database on migration run * Fixed reloading database schema for sqlite by reconnecting and fixing postgres migration * Added checks to Redis and exiting process if connection is unavailable * Fixing error with new installations * Fix issue with data not being sent back to browser on manual executions with defined destination * Merging bull and unify execution id branch fixes * Main process will now get execution success from database instead of redis * Omit execution duration if execution did not stop * Fix issue with execution list displaying inconsistant information information while a workflow is running * Remove unused hooks to clarify for developers that these wont run in queue mode * Added active pooling to help recover from Redis crashes * Lint issues * Changing default polling interval to 60 seconds * Removed unnecessary attributes from bull job * Added webhooks service and setting to disable webhooks from main process * Fixed executions list when running with queues. Now we get the list of actively running workflows from bull. * Add option to disable deregistration of webhooks on shutdown * Rename WEBHOOK_TUNNEL_URL to WEBHOOK_URL keeping backwards compat. * Added auto refresh to executions list * Improvements to workflow stop process when running with queues * Refactor queue system to use a singleton and avoid code duplication * Improve comments and remove unnecessary commits * Remove console.log from vue file * Blocking webhook process to run without queues * Handling execution stop graciously when possible * Removing initialization of all workflows from webhook process * Refactoring code to remove code duplication for job stop * Improved execution list to be more fluid and less intrusive * Fixing workflow name for current executions when auto updating * :zap: Right align autorefresh checkbox Co-authored-by: Jan Oberhauser <jan.oberhauser@gmail.com>
2021-02-09 14:32:40 -08:00
* It is technically possible to run without queues but
* there are 2 known bugs when running in this mode:
* - Executions list will be problematic as the main process
* is not aware of current executions in the webhook processes
* and therefore will display all current executions as error
* as it is unable to determine if it is still running or crashed
* - You cannot stop currently executing jobs from webhook processes
* when running without queues as the main process cannot talk to
* the wehbook processes to communicate workflow execution interruption.
*/
this.error('Webhook processes can only run with execution mode as queue.');
}
try {
// Start directly with the init of the database to improve startup time
const startDbInitPromise = Db.init().catch(error => {
console.error(`There was an error initializing DB: ${error.message}`);
processExistCode = 1;
// @ts-ignore
process.emit('SIGINT');
});
// Make sure the settings exist
const userSettings = await UserSettings.prepareUserSettings();
// Load all node and credential types
const loadNodesAndCredentials = LoadNodesAndCredentials();
await loadNodesAndCredentials.init();
// Load the credentials overwrites if any exist
const credentialsOverwrites = CredentialsOverwrites();
await credentialsOverwrites.init();
// Load all external hooks
const externalHooks = ExternalHooks();
await externalHooks.init();
// Add the found types to an instance other parts of the application can use
const nodeTypes = NodeTypes();
await nodeTypes.init(loadNodesAndCredentials.nodeTypes);
const credentialTypes = CredentialTypes();
await credentialTypes.init(loadNodesAndCredentials.credentialTypes);
// Wait till the database is ready
await startDbInitPromise;
if (config.get('executions.mode') === 'queue') {
const redisHost = config.get('queue.bull.redis.host');
const redisPassword = config.get('queue.bull.redis.password');
const redisPort = config.get('queue.bull.redis.port');
const redisDB = config.get('queue.bull.redis.db');
const redisConnectionTimeoutLimit = config.get('queue.bull.redis.timeoutThreshold');
let lastTimer = 0, cumulativeTimeout = 0;
2021-03-25 03:23:54 -07:00
:sparkles: Separate webhooks from core (#1408) * Unify execution ID across executions * Fix indentation and improved comments * WIP: saving data after each node execution * Added on/off to save data after each step, saving initial data and retries working * Fixing lint issues * Fixing more lint issues * :sparkles: Add bull to execute workflows * :shirt: Fix lint issue * :zap: Add graceful shutdown to worker * :zap: Add loading staticData to worker * :shirt: Fix lint issue * :zap: Fix import * Changed tables metadata to add nullable to stoppedAt * Reload database on migration run * Fixed reloading database schema for sqlite by reconnecting and fixing postgres migration * Added checks to Redis and exiting process if connection is unavailable * Fixing error with new installations * Fix issue with data not being sent back to browser on manual executions with defined destination * Merging bull and unify execution id branch fixes * Main process will now get execution success from database instead of redis * Omit execution duration if execution did not stop * Fix issue with execution list displaying inconsistant information information while a workflow is running * Remove unused hooks to clarify for developers that these wont run in queue mode * Added active pooling to help recover from Redis crashes * Lint issues * Changing default polling interval to 60 seconds * Removed unnecessary attributes from bull job * Added webhooks service and setting to disable webhooks from main process * Fixed executions list when running with queues. Now we get the list of actively running workflows from bull. * Add option to disable deregistration of webhooks on shutdown * Rename WEBHOOK_TUNNEL_URL to WEBHOOK_URL keeping backwards compat. * Added auto refresh to executions list * Improvements to workflow stop process when running with queues * Refactor queue system to use a singleton and avoid code duplication * Improve comments and remove unnecessary commits * Remove console.log from vue file * Blocking webhook process to run without queues * Handling execution stop graciously when possible * Removing initialization of all workflows from webhook process * Refactoring code to remove code duplication for job stop * Improved execution list to be more fluid and less intrusive * Fixing workflow name for current executions when auto updating * :zap: Right align autorefresh checkbox Co-authored-by: Jan Oberhauser <jan.oberhauser@gmail.com>
2021-02-09 14:32:40 -08:00
const settings = {
retryStrategy: (times: number): number | null => {
const now = Date.now();
if (now - lastTimer > 30000) {
// Means we had no timeout at all or last timeout was temporary and we recovered
lastTimer = now;
cumulativeTimeout = 0;
} else {
cumulativeTimeout += now - lastTimer;
lastTimer = now;
if (cumulativeTimeout > redisConnectionTimeoutLimit) {
2021-03-25 03:23:54 -07:00
console.error('Unable to connect to Redis after ' + redisConnectionTimeoutLimit + '. Exiting process.');
:sparkles: Separate webhooks from core (#1408) * Unify execution ID across executions * Fix indentation and improved comments * WIP: saving data after each node execution * Added on/off to save data after each step, saving initial data and retries working * Fixing lint issues * Fixing more lint issues * :sparkles: Add bull to execute workflows * :shirt: Fix lint issue * :zap: Add graceful shutdown to worker * :zap: Add loading staticData to worker * :shirt: Fix lint issue * :zap: Fix import * Changed tables metadata to add nullable to stoppedAt * Reload database on migration run * Fixed reloading database schema for sqlite by reconnecting and fixing postgres migration * Added checks to Redis and exiting process if connection is unavailable * Fixing error with new installations * Fix issue with data not being sent back to browser on manual executions with defined destination * Merging bull and unify execution id branch fixes * Main process will now get execution success from database instead of redis * Omit execution duration if execution did not stop * Fix issue with execution list displaying inconsistant information information while a workflow is running * Remove unused hooks to clarify for developers that these wont run in queue mode * Added active pooling to help recover from Redis crashes * Lint issues * Changing default polling interval to 60 seconds * Removed unnecessary attributes from bull job * Added webhooks service and setting to disable webhooks from main process * Fixed executions list when running with queues. Now we get the list of actively running workflows from bull. * Add option to disable deregistration of webhooks on shutdown * Rename WEBHOOK_TUNNEL_URL to WEBHOOK_URL keeping backwards compat. * Added auto refresh to executions list * Improvements to workflow stop process when running with queues * Refactor queue system to use a singleton and avoid code duplication * Improve comments and remove unnecessary commits * Remove console.log from vue file * Blocking webhook process to run without queues * Handling execution stop graciously when possible * Removing initialization of all workflows from webhook process * Refactoring code to remove code duplication for job stop * Improved execution list to be more fluid and less intrusive * Fixing workflow name for current executions when auto updating * :zap: Right align autorefresh checkbox Co-authored-by: Jan Oberhauser <jan.oberhauser@gmail.com>
2021-02-09 14:32:40 -08:00
process.exit(1);
}
}
return 500;
},
} as IDataObject;
if (redisHost) {
settings.host = redisHost;
}
if (redisPassword) {
settings.password = redisPassword;
}
if (redisPort) {
settings.port = redisPort;
}
if (redisDB) {
settings.db = redisDB;
}
2021-03-25 03:23:54 -07:00
:sparkles: Separate webhooks from core (#1408) * Unify execution ID across executions * Fix indentation and improved comments * WIP: saving data after each node execution * Added on/off to save data after each step, saving initial data and retries working * Fixing lint issues * Fixing more lint issues * :sparkles: Add bull to execute workflows * :shirt: Fix lint issue * :zap: Add graceful shutdown to worker * :zap: Add loading staticData to worker * :shirt: Fix lint issue * :zap: Fix import * Changed tables metadata to add nullable to stoppedAt * Reload database on migration run * Fixed reloading database schema for sqlite by reconnecting and fixing postgres migration * Added checks to Redis and exiting process if connection is unavailable * Fixing error with new installations * Fix issue with data not being sent back to browser on manual executions with defined destination * Merging bull and unify execution id branch fixes * Main process will now get execution success from database instead of redis * Omit execution duration if execution did not stop * Fix issue with execution list displaying inconsistant information information while a workflow is running * Remove unused hooks to clarify for developers that these wont run in queue mode * Added active pooling to help recover from Redis crashes * Lint issues * Changing default polling interval to 60 seconds * Removed unnecessary attributes from bull job * Added webhooks service and setting to disable webhooks from main process * Fixed executions list when running with queues. Now we get the list of actively running workflows from bull. * Add option to disable deregistration of webhooks on shutdown * Rename WEBHOOK_TUNNEL_URL to WEBHOOK_URL keeping backwards compat. * Added auto refresh to executions list * Improvements to workflow stop process when running with queues * Refactor queue system to use a singleton and avoid code duplication * Improve comments and remove unnecessary commits * Remove console.log from vue file * Blocking webhook process to run without queues * Handling execution stop graciously when possible * Removing initialization of all workflows from webhook process * Refactoring code to remove code duplication for job stop * Improved execution list to be more fluid and less intrusive * Fixing workflow name for current executions when auto updating * :zap: Right align autorefresh checkbox Co-authored-by: Jan Oberhauser <jan.oberhauser@gmail.com>
2021-02-09 14:32:40 -08:00
// This connection is going to be our heartbeat
// IORedis automatically pings redis and tries to reconnect
// We will be using the retryStrategy above
// to control how and when to exit.
const redis = new Redis(settings);
redis.on('error', (error) => {
if (error.toString().includes('ECONNREFUSED') === true) {
console.warn('Redis unavailable - trying to reconnect...');
} else {
console.warn('Error with Redis: ', error);
}
});
}
2021-03-25 03:23:54 -07:00
:sparkles: Separate webhooks from core (#1408) * Unify execution ID across executions * Fix indentation and improved comments * WIP: saving data after each node execution * Added on/off to save data after each step, saving initial data and retries working * Fixing lint issues * Fixing more lint issues * :sparkles: Add bull to execute workflows * :shirt: Fix lint issue * :zap: Add graceful shutdown to worker * :zap: Add loading staticData to worker * :shirt: Fix lint issue * :zap: Fix import * Changed tables metadata to add nullable to stoppedAt * Reload database on migration run * Fixed reloading database schema for sqlite by reconnecting and fixing postgres migration * Added checks to Redis and exiting process if connection is unavailable * Fixing error with new installations * Fix issue with data not being sent back to browser on manual executions with defined destination * Merging bull and unify execution id branch fixes * Main process will now get execution success from database instead of redis * Omit execution duration if execution did not stop * Fix issue with execution list displaying inconsistant information information while a workflow is running * Remove unused hooks to clarify for developers that these wont run in queue mode * Added active pooling to help recover from Redis crashes * Lint issues * Changing default polling interval to 60 seconds * Removed unnecessary attributes from bull job * Added webhooks service and setting to disable webhooks from main process * Fixed executions list when running with queues. Now we get the list of actively running workflows from bull. * Add option to disable deregistration of webhooks on shutdown * Rename WEBHOOK_TUNNEL_URL to WEBHOOK_URL keeping backwards compat. * Added auto refresh to executions list * Improvements to workflow stop process when running with queues * Refactor queue system to use a singleton and avoid code duplication * Improve comments and remove unnecessary commits * Remove console.log from vue file * Blocking webhook process to run without queues * Handling execution stop graciously when possible * Removing initialization of all workflows from webhook process * Refactoring code to remove code duplication for job stop * Improved execution list to be more fluid and less intrusive * Fixing workflow name for current executions when auto updating * :zap: Right align autorefresh checkbox Co-authored-by: Jan Oberhauser <jan.oberhauser@gmail.com>
2021-02-09 14:32:40 -08:00
await WebhookServer.start();
// Start to get active workflows and run their triggers
activeWorkflowRunner = ActiveWorkflowRunner.getInstance();
await activeWorkflowRunner.initWebhooks();
const editorUrl = GenericHelpers.getBaseUrl();
this.log('Webhook listener waiting for requests.');
} catch (error) {
this.error(`There was an error: ${error.message}`);
processExistCode = 1;
// @ts-ignore
process.emit('SIGINT');
}
})();
}
}