sourcebot/packages/backend/src/connectionManager.ts
Brendan Kellam 39b92b9e98
v3 effort (#158)
* SQL Database (#157)

* point zoekt to v3 branch

* bump zoekt version

* Add tenant ID concept into web app and backend (#160)

* hacked together a example of using zoekt grpc api

* provide tenant id to zoekt git indexer

* update zoekt version to point to multitenant branch

* pipe tenant id through header to zoekt

* remove incorrect submodule reference and settings typo

* update zoekt commit

* remove unused yarn script

* remove unused grpc client in web server

* remove unneeded deps and improve tenant id log

* pass tenant id when creating repo in db

* add mt yarn script

* add nocheckin comment to tenant id in v2 schema

---------

Co-authored-by: bkellam <bshizzle1234@gmail.com>

* bump zoekt version

* parallelize repo indexing (#163)

* hacked together a example of using zoekt grpc api

* provide tenant id to zoekt git indexer

* update zoekt version to point to multitenant branch

* pipe tenant id through header to zoekt

* remove incorrect submodule reference and settings typo

* update zoekt commit

* remove unused yarn script

* remove unused grpc client in web server

* remove unneeded deps and improve tenant id log

* pass tenant id when creating repo in db

* add mt yarn script

* add pol of bullmq into backend

* add better error handling and concurrency setting

* spin up redis instance in dockerfile

* cleanup transaction logic when adding repos to index queue

* add NEW index status fetch condition

* move bullmq deps to backend

---------

Co-authored-by: bkellam <bshizzle1234@gmail.com>

* Authentication (#164)

* Add Org table (#167)

* Move logout button & profile picture into settings dropdown (#172)

* Multi tenancy support in config syncer (#171)

* [wip] initial mt support in config syncer

* Move logout button & profile picture into settings dropdown (#172)

* update sync status properly and fix bug with multiple config in db case

* make config path required in single tenant mode

NOTE: deleting config/repos is currently not supported in multi tenancy case. Support for this will be added in a future PR

---------

Co-authored-by: Brendan Kellam <bshizzle1234@gmail.com>

* add tenant mode support in docker container:

* Organization switching & active org management (#173)

* updated syncedAt date after config sync:

* Migrate to postgres (#174)

* spin up postgres in docker container

* get initial pol of postgres db working in docker image

* spin up postgres server in dev case

* updated syncedAt date after config sync:

* remove unnecessary port expose in docker file

* Connection creation form (#175)

* fix issue with yarn dev startup

* init (#176)

* Add `@sourcebot/schemas` package (#177)

* Connection management (#178)

* add concept of secrets (#180)

* add @sourcebot/schemas package

* migrate things to use the schemas package

* Dockerfile support

* add secret table to schema

* Add concept of connection manager

* Rename Config->Connection

* Handle job failures

* Add join table between repo and connection

* nits

* create first version of crypto package

* add crypto package as deps to others

* forgot to add package changes

* add server action for adding and listing secrets, create test page for it

* add secrets page to nav menu

* add secret to config and support fetching it in backend

* reset secret form on successful submission

* add toast feedback for secrets form

* add instructions for adding encryption key to dev instructions

* add encryption key support in docker file

* add delete secret button

* fix nits from pr review

---------

Co-authored-by: bkellam <bshizzle1234@gmail.com>

* bump zoekt version

* enforce tenancy on search and repo listing endpoints (#181)

* enforce tenancy on search and repo listing

* remove orgId from request schemas

* adds garbage collection for repos (#182)

* refactor repo indexing logic into RepoManager

* wip cleanup stale repos

* add rest of gc logic

* set status to indexing properly

* add initial logic for staging environment

* try to move encryption key env decleration in docker file to fix build issues

* switch encryption key as build arg to se if that fixes build issues

* add deployment action for staging image

* try using mac github action runners instead

* switch to using arm64 runners on arm64 build

* change workflow names to fix trigger issue

* trigger staging actions to see if it works

* fix working directory typo and pray it doesnt push to prod

* checkout v3 when deploying staging

* try to change into the staging dir manuall

* dummy commit to trigger v3 workflows to test

* update staging deploy script to match new version in main

* reference proper image:tag in staging fly config

* update staging fly config to point to ghcr

* Connection management (#183)

* add invite system and google oauth provider (#185)

* add settings page with members list

* add invite to schema and basic create form

* add invite table

* add basic invite link copy button

* add auth invite accept case

* add non auth logic

* add google oauth provider

* fix reference to header component in connections

* add google logo to google oauth

* fix web build errors

* bump staging resources

* change staging cpu to perf

* add side bar nav in settings page

* improve styling of members page

* wip adding stripe checkout button

* wip onboarding flow

* add stripe subscription id to org

* save stripe session id and add manage subscription button in settings

* properly block access to pages if user isn't in an org

* wip add paywall

* Domain support

* Domain support (#188)

* Update Makefile to include crypto package when doing a make clean

* Add default for AUTH_URL in attempt to fix build

* attempt 2

* fix attempt #3: Do not require a encrpytion key at build time

* Fix generate script race condition

* Attempt #4

* add back paywall and also add support for incrememnting seat count on invite redemption

* prevent self invite

* action button styling in settings and toast on copy

* add ability to remove member from org

* move stripe product id to env var

* add await for blocking loop in backend

* add subscription info to billing page

* handle trial case in billing info page

* add trial duration indicator to nav bar

* check if domain starts or ends with dash

* remove unused no org component

* Generate AUTH_SECRET if not provided (#189)

* remove package lock file and fix prisma dep version

* revert dep version updates

* fix yarn.lock

* add auth and membership check to fetchSubscription

* properly handle invite redeem with no valid subscription case

* change back fetch subscription to not require org membership

* add back subscription check in invite redeem page

* Add stripe billing logic (#190)

* add side bar nav in settings page

* improve styling of members page

* wip adding stripe checkout button

* wip onboarding flow

* add stripe subscription id to org

* save stripe session id and add manage subscription button in settings

* properly block access to pages if user isn't in an org

* wip add paywall

* Domain support

* add back paywall and also add support for incrememnting seat count on invite redemption

* prevent self invite

* action button styling in settings and toast on copy

* add ability to remove member from org

* move stripe product id to env var

* add await for blocking loop in backend

* add subscription info to billing page

* handle trial case in billing info page

* add trial duration indicator to nav bar

* check if domain starts or ends with dash

* remove unused no org component

* remove package lock file and fix prisma dep version

* revert dep version updates

* fix yarn.lock

* add auth and membership check to fetchSubscription

* properly handle invite redeem with no valid subscription case

* change back fetch subscription to not require org membership

* add back subscription check in invite redeem page

---------

Co-authored-by: bkellam <bshizzle1234@gmail.com>

* fix nits

* remove providers check

* fix more nits

* change stripe init to be behind function

* fix publishible stripe key handling in docker container

* enforce owner perms (#191)

* add make owner logic, and owner perms for removal, invite, and manage subscription

* add change billing email card to billing settings

* enforce owner role in action level

* remove unused hover card component

* cleanup

* add back gitlab, gitea, and gerrit support (#184)

* add non github config definitions

* refactor github config compilation to seperate file

* add gitlab config compilation

* Connection management (#183)

* wip gitlab repo sync support

* fix gitlab zoekt metadata

* add gitea support

* add gerrit support

* Connection management (#183)

* add gerrit config compilation

* Connection management (#183)

---------

Co-authored-by: Brendan Kellam <bshizzle1234@gmail.com>

* fix apos usage in redeem page

* change csrf cookie to secure not host

* Credentials provider (#192)

* email password functionality

* feedback

* cleanup org's repos and shards if it's inactive (#194)

* add stripe subscription status and webhook

* add inactive org repo cleanup logic

* mark reactivated org connections for sync

* connections qol improvements (#195)

* add client side polling to connections list

* properly fetch repo image url

* add client polling to connection management page, and add ability to sync failed connections

* Fix build with suspense boundary

* improved fix

* add retries for 429 issues (#196)

* add connection compile retry and hard repo limit

* add more retry checks

* cleanup unused change

* address feedback

* fix build errors and add index concurrency env var

* add config upsert timeout env var

* Membership settings rework (#198)

* Add refined members list

* futher progress on members settings polish

* Remove old components

* feedback

* Magic links (#199)

* wip on magic link support

* Switch to nodemailer / resend for transactional mail

* Further cleanup

* Add stylized email using react-email

* fix

* Fix build

* db performance improvements and job resilience  (#200)

* replace upsert with seperate create many and raw update many calls

* add bulk repo status update and queue addition with priority

* add support for managed redis

* add note for changing raw sql on schema change

* remove non secret token options

* fix token examples in schema

* add better visualization for connection/repo errors and warnings (#201)

* replace upsert with seperate create many and raw update many calls

* add bulk repo status update and queue addition with priority

* add support for managed redis

* add note for changing raw sql on schema change

* add error package and use BackendException in connection manager

* handle connection failure display on web app

* add warning banner for not found orgs/repos/users

* add failure handling for gerrit

* add gitea notfound warning support

* add warning icon in connections list

* style nits

* add failed repo vis in connections list

* added retry failed repo index buttons

* move nav indicators to client with polling

* fix indicator flash issue and truncate large list results

* display error nav better

* truncate failed repo list in connection list item

* fix merge error

* fix merge bug

* add connection util file [wip]

* refactor notfound fetch logic and add missing error package to dockerfile

* move repeated logic to function and add zod schema for syncStatusMetadata

* add orgid unique constraint to repo

* revert repo compile update logic to upsert loop

* log upsert stats

* [temp] disable polling everywhere (#205)

* add health check endpoint

* Refined onboarding flow (#202)

* Redeem UX pass (#204)

* add log for health check

* fix new connection complete callback route

* add cpu split logic and only wait for postgres if we're going to connec to it

* Inline secret creation (#207)

* use docker scopes to try and improve caching

* Dummy change

* remove cpu split logic

* Add some instrumentation to web

* add posthog events on various user actions (#208)

* add page view event support

* add posthog events

* nit: remove unused import

* feedback

* fix merge error

* use staging posthog papik when building staging image

* fix other merge error and build warnings

* Add invite email (#209)

* wrap posthog provider in suspense to fix build error

* add grafana alloy config and setup (#210)

* add grafana alloy config and setup

* add basic repo prom metrics

* nits in dockerfile

* remove invalid characters when auto filling domain

* add login posthog events

* remove hard coded sourcebot.app references

* make repo garbage collection async (#211)

* add gc queue logic

* fix missing switch cases for gc status

* style org create form better with new staging domain

* change repo rm logic to be async

* simplify repo for inactive org query

* add grace period for garbage collecting repos

* make prom scrape interval 500ms

* fix typo in trial card

* onboarding tweaks

* rename some prom metrics and cleanup unused

* wipe existing repo if we've picked up a killed job to ensure good state

* Connections UX pass + query optimizations (#212)

* remove git & local schemas (#213)

* skip stripe checkout for trial + fix indexing in progress UI + additional schema validation (#214)

* add additional config validation

* wip bypass stripe checkout for trial

* fix stripe trial checkout bypass

* fix indexing in progress ui on home page

* add subscription checks, more schema validation, and fix issue with complete page

* dont display if no indexed repos

* fix skipping onboard complete check

* fix build error

* add back button in onboard connection creation flow

* Add back revision support (#215)

* fix build

* Fix bug with repository snapshot

* fix share links

* fix repo rm issue, 502 page, condition on test clock

* Make login and onboarding mobile friendly

* fix ordering of quick actions

* remove error msg dump on failed repo index job, and update indexedAt field

* Add mobile unsupported splash screne

* cherry pick fix for file links

* [Cherry Pick] Syntax reference guide (#169) (#216)

* Add .env to db gitignore

* fix case where we have repos but they're all failed for repo snapshot

* /settings/secrets page (#217)

* display domain properly in org create form

* Quick action tweaks (#218)

* revamp repo page (#220)

* wip repo table

* new repo page

* add indicator for when feedback is applied in repo page

* add repo button

* fetch connection data in one query

* fix styling

* fix (#219)

* remove / keyboard shortcut hint in search bar

* prevent switching to first page on data update and truncate long repo names in repo list

* General settings + cleanup (#221)

* General settings

* Add alert to org domain change

* First attempt at sending logs to grafana

* logs wip

* add alloy logs

* wip

* [temp] comment out loki for now

* update trial card content and add events for code host selection on onboard

* reduce scraping interval to 15s

* Add prometheus metric for pending repo indexing jobs

* switch magic link to invite code (#222)

* wip magic link codes

* pipe email to email provider properly

* remove magic link data cookie after sign in

* clean up unused imports

* dont remove cookie before we use it

* rm package-lock.json

* revert yarn files to v3 state

* switch email passing from cookie to search param

* add comment for settings dropdown auth update

* remove unused middleware file

* fix build error and warnings

* fix build error with useSearchParam not wrapped in suspense

* add sentry support to backend and webapp (#223)

* add sentry to web app

* set sentry environemnt from env var

* add sentry env replace logic in docker container

* wip add backend sentry

* add sentry to backend

* move dns to env var

* remove test exception

* Fix root domain issue on onboarding

* add setup sentry cli step to github action

* login to sentry

* fix sentry login in action

* Update grafana loki endpoint

* switch source map publish to runtime in entrypoint

* catch and rethrow simplegit exceptions

* alloy nits

* fix alloy

* backend logging (#224)

* revert grafana loki config

* fix login ui nits

* fix quick actions

* fix typo in secret creation

* fix private repo clone issue for gitlab

* add repo index timeout logic

* add posthog identify call after registeration

* various changes to add terms and security info (#225)

* add terms and security to footer

* add security card

* add demo card

* fix build error

* nit fix: center 'get in touch' on security card

* Dark theme improvements (#226)

* (fix) Fixed bug with gitlab and gitea not including hostname in the repoName

* Switch to using t3-env for env-var management (#230)

* Add missing env var

* fix build

* Centralize to using a single .env.development for development workflows (#231)

* Make billing optional (#232)

* Massage environment variables from strings to numbers (#234)

* Single tenancy & auth modes (#233)

* Add docs to this repo

* dummy change

* Declarative connection configuration (#235)

* fix build

* upgrade to next 14.2.25

* Improved database DX

* migrate to yarn v4

* Use origin from header for baseUrl of emails (instead of AUTH_URL). Also removed reference to hide scrollbars

* Remove SOURCEBOT_ENCRYPTION_KEY from build arg

* Fix issue with linking default user to org in single tenant + no-auth mode

* Fix fallback tokens (#242)

* add SECURITY_CARD_ENABLED flag

* Add repository weburl (#243)

* Random fixes and improvements (#244)

* add zoekt max wall time env var

* remove empty warning in docs

* fix reference in sh docs

* add connection manager upsert timeout env var

* Declarative connection cleanup + improvements (#245)

* change contact us footer in app to point to main contact form

* PostHog event pass (#246)

* fix typo

* Add sourcebot cloud environment prop to staging workflow

* Update generated files

* remove AUTH_URL since it unused and (likely) unnecessary

* Revert "remove AUTH_URL since it unused and (likely) unnecessary"

This reverts commit 1f4a5aed22.

* cleanup GitHub action releases (#252)

* remove alloy, change auth defaul to disabled, add settings page in me dropdown

* enforce connection management perms to owner (#253)

* enforce conneciton management perms to owner

* fix formatting

* more formatting

* naming nits

* fix var name error

* change empty repo set copy if auth is disabled

* add CONTRIBUTING.md file

* hide settings in dropdown with auth isnt enabled

* handle case where gerrit weburl is just gitiles path

* Docs overhall (#251)

* remove nocheckin

* fix build error

* remove v3 trigger from deploy staging

* fix build errors round 2

* another error fix

---------

Co-authored-by: msukkari <michael.sukkarieh@mail.mcgill.ca>
2025-03-31 22:34:42 -07:00

318 lines
12 KiB
TypeScript

import { Connection, ConnectionSyncStatus, PrismaClient, Prisma } from "@sourcebot/db";
import { Job, Queue, Worker } from 'bullmq';
import { Settings } from "./types.js";
import { ConnectionConfig } from "@sourcebot/schemas/v3/connection.type";
import { createLogger } from "./logger.js";
import { Redis } from 'ioredis';
import { RepoData, compileGithubConfig, compileGitlabConfig, compileGiteaConfig, compileGerritConfig } from "./repoCompileUtils.js";
import { BackendError, BackendException } from "@sourcebot/error";
import { captureEvent } from "./posthog.js";
import { env } from "./env.js";
import * as Sentry from "@sentry/node";
interface IConnectionManager {
scheduleConnectionSync: (connection: Connection) => Promise<void>;
registerPollingCallback: () => void;
dispose: () => void;
}
const QUEUE_NAME = 'connectionSyncQueue';
type JobPayload = {
connectionId: number,
orgId: number,
config: ConnectionConfig,
};
type JobResult = {
repoCount: number,
}
export class ConnectionManager implements IConnectionManager {
private worker: Worker;
private queue: Queue<JobPayload>;
private logger = createLogger('ConnectionManager');
constructor(
private db: PrismaClient,
private settings: Settings,
redis: Redis,
) {
this.queue = new Queue<JobPayload>(QUEUE_NAME, {
connection: redis,
});
this.worker = new Worker(QUEUE_NAME, this.runSyncJob.bind(this), {
connection: redis,
concurrency: this.settings.maxConnectionSyncJobConcurrency,
});
this.worker.on('completed', this.onSyncJobCompleted.bind(this));
this.worker.on('failed', this.onSyncJobFailed.bind(this));
}
public async scheduleConnectionSync(connection: Connection) {
await this.db.$transaction(async (tx) => {
await tx.connection.update({
where: { id: connection.id },
data: { syncStatus: ConnectionSyncStatus.IN_SYNC_QUEUE },
});
const connectionConfig = connection.config as unknown as ConnectionConfig;
await this.queue.add('connectionSyncJob', {
connectionId: connection.id,
orgId: connection.orgId,
config: connectionConfig,
});
this.logger.info(`Added job to queue for connection ${connection.id}`);
}).catch((err: unknown) => {
this.logger.error(`Failed to add job to queue for connection ${connection.id}: ${err}`);
});
}
public async registerPollingCallback() {
setInterval(async () => {
const connections = await this.db.connection.findMany({
where: {
syncStatus: ConnectionSyncStatus.SYNC_NEEDED,
}
});
for (const connection of connections) {
await this.scheduleConnectionSync(connection);
}
}, this.settings.resyncConnectionPollingIntervalMs);
}
private async runSyncJob(job: Job<JobPayload>): Promise<JobResult> {
const { config, orgId } = job.data;
// @note: We aren't actually doing anything with this atm.
const abortController = new AbortController();
const connection = await this.db.connection.findUnique({
where: {
id: job.data.connectionId,
},
});
if (!connection) {
const e = new BackendException(BackendError.CONNECTION_SYNC_CONNECTION_NOT_FOUND, {
message: `Connection ${job.data.connectionId} not found`,
});
Sentry.captureException(e);
throw e;
}
// Reset the syncStatusMetadata to an empty object at the start of the sync job
await this.db.connection.update({
where: {
id: job.data.connectionId,
},
data: {
syncStatus: ConnectionSyncStatus.SYNCING,
syncStatusMetadata: {}
}
})
let result: {
repoData: RepoData[],
notFound: {
users: string[],
orgs: string[],
repos: string[],
}
} = {
repoData: [],
notFound: {
users: [],
orgs: [],
repos: [],
}
};
try {
result = await (async () => {
switch (config.type) {
case 'github': {
return await compileGithubConfig(config, job.data.connectionId, orgId, this.db, abortController);
}
case 'gitlab': {
return await compileGitlabConfig(config, job.data.connectionId, orgId, this.db);
}
case 'gitea': {
return await compileGiteaConfig(config, job.data.connectionId, orgId, this.db);
}
case 'gerrit': {
return await compileGerritConfig(config, job.data.connectionId, orgId);
}
}
})();
} catch (err) {
this.logger.error(`Failed to compile repo data for connection ${job.data.connectionId}: ${err}`);
Sentry.captureException(err);
if (err instanceof BackendException) {
throw err;
} else {
throw new BackendException(BackendError.CONNECTION_SYNC_SYSTEM_ERROR, {
message: `Failed to compile repo data for connection ${job.data.connectionId}`,
});
}
}
let { repoData, notFound } = result;
// Push the information regarding not found users, orgs, and repos to the connection's syncStatusMetadata. Note that
// this won't be overwritten even if the connection job fails
await this.db.connection.update({
where: {
id: job.data.connectionId,
},
data: {
syncStatusMetadata: { notFound }
}
});
// Filter out any duplicates by external_id and external_codeHostUrl.
repoData = repoData.filter((repo, index, self) => {
return index === self.findIndex(r =>
r.external_id === repo.external_id &&
r.external_codeHostUrl === repo.external_codeHostUrl
);
})
// @note: to handle orphaned Repos we delete all RepoToConnection records for this connection,
// and then recreate them when we upsert the repos. For example, if a repo is no-longer
// captured by the connection's config (e.g., it was deleted, marked archived, etc.), it won't
// appear in the repoData array above, and so the RepoToConnection record won't be re-created.
// Repos that have no RepoToConnection records are considered orphaned and can be deleted.
await this.db.$transaction(async (tx) => {
const deleteStart = performance.now();
await tx.connection.update({
where: {
id: job.data.connectionId,
},
data: {
repos: {
deleteMany: {}
}
}
});
const deleteDuration = performance.now() - deleteStart;
this.logger.info(`Deleted all RepoToConnection records for connection ${job.data.connectionId} in ${deleteDuration}ms`);
const totalUpsertStart = performance.now();
for (const repo of repoData) {
const upsertStart = performance.now();
await tx.repo.upsert({
where: {
external_id_external_codeHostUrl_orgId: {
external_id: repo.external_id,
external_codeHostUrl: repo.external_codeHostUrl,
orgId: orgId,
}
},
update: repo,
create: repo,
})
const upsertDuration = performance.now() - upsertStart;
this.logger.info(`Upserted repo ${repo.external_id} in ${upsertDuration}ms`);
}
const totalUpsertDuration = performance.now() - totalUpsertStart;
this.logger.info(`Upserted ${repoData.length} repos in ${totalUpsertDuration}ms`);
}, { timeout: env.CONNECTION_MANAGER_UPSERT_TIMEOUT_MS });
return {
repoCount: repoData.length,
};
}
private async onSyncJobCompleted(job: Job<JobPayload>, result: JobResult) {
this.logger.info(`Connection sync job ${job.id} completed`);
const { connectionId } = job.data;
let syncStatusMetadata: Record<string, unknown> = (await this.db.connection.findUnique({
where: { id: connectionId },
select: { syncStatusMetadata: true }
}))?.syncStatusMetadata as Record<string, unknown> ?? {};
const { notFound } = syncStatusMetadata as { notFound: {
users: string[],
orgs: string[],
repos: string[],
}};
await this.db.connection.update({
where: {
id: connectionId,
},
data: {
syncStatus:
notFound.users.length > 0 ||
notFound.orgs.length > 0 ||
notFound.repos.length > 0 ? ConnectionSyncStatus.SYNCED_WITH_WARNINGS : ConnectionSyncStatus.SYNCED,
syncedAt: new Date()
}
})
captureEvent('backend_connection_sync_job_completed', {
connectionId: connectionId,
repoCount: result.repoCount,
});
}
private async onSyncJobFailed(job: Job<JobPayload> | undefined, err: unknown) {
this.logger.info(`Connection sync job failed with error: ${err}`);
Sentry.captureException(err, {
tags: {
connectionid: job?.data.connectionId,
jobId: job?.id,
queue: QUEUE_NAME,
}
});
if (job) {
const { connectionId } = job.data;
captureEvent('backend_connection_sync_job_failed', {
connectionId: connectionId,
error: err instanceof BackendException ? err.code : 'UNKNOWN',
});
// We may have pushed some metadata during the execution of the job, so we make sure to not overwrite the metadata here
let syncStatusMetadata: Record<string, unknown> = (await this.db.connection.findUnique({
where: { id: connectionId },
select: { syncStatusMetadata: true }
}))?.syncStatusMetadata as Record<string, unknown> ?? {};
if (err instanceof BackendException) {
syncStatusMetadata = {
...syncStatusMetadata,
error: err.code,
...err.metadata,
}
} else {
syncStatusMetadata = {
...syncStatusMetadata,
error: 'UNKNOWN',
}
}
await this.db.connection.update({
where: {
id: connectionId,
},
data: {
syncStatus: ConnectionSyncStatus.FAILED,
syncedAt: new Date(),
syncStatusMetadata: syncStatusMetadata as Prisma.InputJsonValue,
}
});
}
}
public dispose() {
this.worker.close();
this.queue.close();
}
}