You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
gristlabs_grist-core/app/server/lib/DocApi.ts

701 lines
30 KiB

import { createEmptyActionSummary } from "app/common/ActionSummary";
import { ApiError } from 'app/common/ApiError';
import { BrowserSettings } from "app/common/BrowserSettings";
import { fromTableDataAction, TableColValues } from 'app/common/DocActions';
import { arrayRepeat, isAffirmative } from "app/common/gutil";
import { SortFunc } from 'app/common/SortFunc';
import { DocReplacementOptions, DocState, DocStateComparison, DocStates, NEW_DOCUMENT_CODE} from 'app/common/UserAPI';
import { HomeDBManager, makeDocAuthResult } from 'app/gen-server/lib/HomeDBManager';
import { concatenateSummaries, summarizeAction } from "app/server/lib/ActionSummary";
import { ActiveDoc } from "app/server/lib/ActiveDoc";
import { assertAccess, getOrSetDocAuth, getTransitiveHeaders, getUserId, isAnonymousUser,
RequestWithLogin } from 'app/server/lib/Authorizer';
import { DocManager } from "app/server/lib/DocManager";
import { docSessionFromRequest, makeExceptionalDocSession, OptDocSession } from "app/server/lib/DocSession";
import { DocWorker } from "app/server/lib/DocWorker";
(core) support GRIST_WORKER_GROUP to place worker into an exclusive group Summary: In an emergency, we may want to serve certain documents with "old" workers as we fix problems. This diff adds some support for that. * Creates duplicate task definitions and services for staging and production doc workers (called grist-docs-staging2 and grist-docs-prod2), pulling from distinct docker tags (staging2 and prod2). The services are set to have zero workers until we need them. * These new workers are started with a new env variable `GRIST_WORKER_GROUP` set to `secondary`. * The `GRIST_WORKER_GROUP` variable, if set, makes the worker available to documents in the named group, and only that group. * An unauthenticated `/assign` endpoint is added to documents which, when POSTed to, checks that the doc is served by a worker in the desired group for that doc (as set manually in redis), and if not frees the doc up for reassignment. This makes it possible to move individual docs between workers without redeployments. The bash scripts added are a record of how the task definitions + services were created. The services could just have been copied manually, but the task definitions will need to be updated whenever the definitions for the main doc workers are updated, so it is worth scripting that. For example, if a certain document were to fail on a new deployment of Grist, but rolling back the full deployment wasn't practical: * Set prod2 tag in docker to desired codebase for that document * Set desired_count for grist-docs-prod2 service to non-zero * Set doc-<docid>-group for that doc in redis to secondary * Hit /api/docs/<docid>/assign to move the doc to grist-docs-prod2 (If the document needs to be reverted to a previous snapshot, that currently would need doing manually - could be made simpler, but not in scope of this diff). Test Plan: added tests Reviewers: dsagal Reviewed By: dsagal Differential Revision: https://phab.getgrist.com/D2649
4 years ago
import { IDocWorkerMap } from "app/server/lib/DocWorkerMap";
import { expressWrap } from 'app/server/lib/expressWrap';
import { GristServer } from 'app/server/lib/GristServer';
import { HashUtil } from 'app/server/lib/HashUtil';
import { makeForkIds } from "app/server/lib/idUtils";
import { getDocId, getDocScope, integerParam, isParameterOn, optStringParam,
sendOkReply, sendReply, stringParam } from 'app/server/lib/requestUtils';
import { SandboxError } from "app/server/lib/sandboxUtil";
import { handleOptionalUpload, handleUpload } from "app/server/lib/uploads";
import * as contentDisposition from 'content-disposition';
import { Application, NextFunction, Request, RequestHandler, Response } from "express";
import fetch from 'node-fetch';
import * as path from 'path';
// Cap on the number of requests that can be outstanding on a single document via the
// rest doc api. When this limit is exceeded, incoming requests receive an immediate
// reply with status 429.
const MAX_PARALLEL_REQUESTS_PER_DOC = 10;
type WithDocHandler = (activeDoc: ActiveDoc, req: RequestWithLogin, resp: Response) => Promise<void>;
/**
* Middleware to track the number of requests outstanding on each document, and to
* throw an exception when the maximum number of requests are already outstanding.
* Access to a document must already have been authorized.
*/
function apiThrottle(usage: Map<string, number>,
callback: (req: RequestWithLogin,
resp: Response,
next: NextFunction) => void|Promise<void>): RequestHandler {
return async (req, res, next) => {
const docId = getDocId(req);
try {
const count = usage.get(docId) || 0;
usage.set(docId, count + 1);
if (count + 1 > MAX_PARALLEL_REQUESTS_PER_DOC) {
throw new ApiError(`Too many backlogged requests for document ${docId} - ` +
`try again later?`, 429);
}
await callback(req as RequestWithLogin, res, next);
} catch (err) {
next(err);
} finally {
const count = usage.get(docId);
if (count) {
if (count === 1) {
usage.delete(docId);
} else {
usage.set(docId, count - 1);
}
}
}
};
}
export class DocWorkerApi {
(core) support GRIST_WORKER_GROUP to place worker into an exclusive group Summary: In an emergency, we may want to serve certain documents with "old" workers as we fix problems. This diff adds some support for that. * Creates duplicate task definitions and services for staging and production doc workers (called grist-docs-staging2 and grist-docs-prod2), pulling from distinct docker tags (staging2 and prod2). The services are set to have zero workers until we need them. * These new workers are started with a new env variable `GRIST_WORKER_GROUP` set to `secondary`. * The `GRIST_WORKER_GROUP` variable, if set, makes the worker available to documents in the named group, and only that group. * An unauthenticated `/assign` endpoint is added to documents which, when POSTed to, checks that the doc is served by a worker in the desired group for that doc (as set manually in redis), and if not frees the doc up for reassignment. This makes it possible to move individual docs between workers without redeployments. The bash scripts added are a record of how the task definitions + services were created. The services could just have been copied manually, but the task definitions will need to be updated whenever the definitions for the main doc workers are updated, so it is worth scripting that. For example, if a certain document were to fail on a new deployment of Grist, but rolling back the full deployment wasn't practical: * Set prod2 tag in docker to desired codebase for that document * Set desired_count for grist-docs-prod2 service to non-zero * Set doc-<docid>-group for that doc in redis to secondary * Hit /api/docs/<docid>/assign to move the doc to grist-docs-prod2 (If the document needs to be reverted to a previous snapshot, that currently would need doing manually - could be made simpler, but not in scope of this diff). Test Plan: added tests Reviewers: dsagal Reviewed By: dsagal Differential Revision: https://phab.getgrist.com/D2649
4 years ago
constructor(private _app: Application, private _docWorker: DocWorker,
private _docWorkerMap: IDocWorkerMap, private _docManager: DocManager,
private _dbManager: HomeDBManager, private _grist: GristServer) {}
/**
* Adds endpoints for the doc api.
*
* Note that it expects bodyParser, userId, and jsonErrorHandler middleware to be set up outside
* to apply to these routes.
*/
public addEndpoints() {
// check document exists (not soft deleted) and user can view it
const canView = expressWrap(this._assertAccess.bind(this, 'viewers', false));
// check document exists (not soft deleted) and user can edit it
const canEdit = expressWrap(this._assertAccess.bind(this, 'editors', false));
// check user can edit document, with soft-deleted documents being acceptable
const canEditMaybeRemoved = expressWrap(this._assertAccess.bind(this, 'editors', true));
(core) support GRIST_WORKER_GROUP to place worker into an exclusive group Summary: In an emergency, we may want to serve certain documents with "old" workers as we fix problems. This diff adds some support for that. * Creates duplicate task definitions and services for staging and production doc workers (called grist-docs-staging2 and grist-docs-prod2), pulling from distinct docker tags (staging2 and prod2). The services are set to have zero workers until we need them. * These new workers are started with a new env variable `GRIST_WORKER_GROUP` set to `secondary`. * The `GRIST_WORKER_GROUP` variable, if set, makes the worker available to documents in the named group, and only that group. * An unauthenticated `/assign` endpoint is added to documents which, when POSTed to, checks that the doc is served by a worker in the desired group for that doc (as set manually in redis), and if not frees the doc up for reassignment. This makes it possible to move individual docs between workers without redeployments. The bash scripts added are a record of how the task definitions + services were created. The services could just have been copied manually, but the task definitions will need to be updated whenever the definitions for the main doc workers are updated, so it is worth scripting that. For example, if a certain document were to fail on a new deployment of Grist, but rolling back the full deployment wasn't practical: * Set prod2 tag in docker to desired codebase for that document * Set desired_count for grist-docs-prod2 service to non-zero * Set doc-<docid>-group for that doc in redis to secondary * Hit /api/docs/<docid>/assign to move the doc to grist-docs-prod2 (If the document needs to be reverted to a previous snapshot, that currently would need doing manually - could be made simpler, but not in scope of this diff). Test Plan: added tests Reviewers: dsagal Reviewed By: dsagal Differential Revision: https://phab.getgrist.com/D2649
4 years ago
// check document exists, don't check user access
const docExists = expressWrap(this._assertAccess.bind(this, null, false));
// Middleware to limit number of outstanding requests per document. Will also
// handle errors like expressWrap would.
const throttled = apiThrottle.bind(null, new Map());
const withDoc = (callback: WithDocHandler) => throttled(this._requireActiveDoc(callback));
// Apply user actions to a document.
this._app.post('/api/docs/:docId/apply', canEdit, withDoc(async (activeDoc, req, res) => {
res.json(await activeDoc.applyUserActions(docSessionFromRequest(req), req.body));
}));
// Get the specified table.
this._app.get('/api/docs/:docId/tables/:tableId/data', canView, withDoc(async (activeDoc, req, res) => {
const filters = req.query.filter ? JSON.parse(String(req.query.filter)) : {};
if (!Object.keys(filters).every(col => Array.isArray(filters[col]))) {
throw new ApiError("Invalid query: filter values must be arrays", 400);
}
const tableId = req.params.tableId;
const tableData = await handleSandboxError(tableId, [], activeDoc.fetchQuery(
docSessionFromRequest(req), {tableId, filters}, true));
// Apply sort/limit parameters, if set. TODO: move sorting/limiting into data engine
// and sql.
const params = getQueryParameters(req);
res.json(applyQueryParameters(fromTableDataAction(tableData), params));
}));
// The upload should be a multipart post with an 'upload' field containing one or more files.
// Returns the list of rowIds for the rows created in the _grist_Attachments table.
this._app.post('/api/docs/:docId/attachments', canEdit, withDoc(async (activeDoc, req, res) => {
const uploadResult = await handleUpload(req, res);
res.json(await activeDoc.addAttachments(docSessionFromRequest(req), uploadResult.uploadId));
}));
// Returns the metadata for a given attachment ID (i.e. a rowId in _grist_Attachments table).
this._app.get('/api/docs/:docId/attachments/:attId', canView, withDoc(async (activeDoc, req, res) => {
const attRecord = activeDoc.getAttachmentMetadata(req.params.attId as string);
const {fileName, fileSize, timeUploaded: t} = attRecord;
const timeUploaded = (typeof t === 'number') ? new Date(t).toISOString() : undefined;
res.json({fileName, fileSize, timeUploaded});
}));
// Responds with attachment contents, with suitable Content-Type and Content-Disposition.
this._app.get('/api/docs/:docId/attachments/:attId/download', canView, withDoc(async (activeDoc, req, res) => {
const attRecord = activeDoc.getAttachmentMetadata(req.params.attId as string);
const fileIdent = attRecord.fileIdent as string;
const ext = path.extname(fileIdent);
const origName = attRecord.fileName as string;
const fileName = ext ? path.basename(origName, path.extname(origName)) + ext : origName;
const fileData = await activeDoc.getAttachmentData(docSessionFromRequest(req), fileIdent);
res.status(200)
.type(ext)
// Construct a content-disposition header of the form 'attachment; filename="NAME"'
.set('Content-Disposition', contentDisposition(fileName, {type: 'attachment'}))
.set('Cache-Control', 'private, max-age=3600')
.send(fileData);
}));
// Adds records.
this._app.post('/api/docs/:docId/tables/:tableId/data', canEdit, withDoc(async (activeDoc, req, res) => {
const tableId = req.params.tableId;
const columnValues = req.body;
const colNames = Object.keys(columnValues);
// user actions expect [null, ...] as row ids, first let's figure the number of items to add by
// looking at the length of a column
const count = columnValues[colNames[0]].length;
// then, let's create [null, ...]
const rowIds = arrayRepeat(count, null);
const sandboxRes = await handleSandboxError(tableId, colNames, activeDoc.applyUserActions(
docSessionFromRequest(req),
[['BulkAddRecord', tableId, rowIds, columnValues]]));
res.json(sandboxRes.retValues[0]);
}));
this._app.post('/api/docs/:docId/tables/:tableId/data/delete', canEdit, withDoc(async (activeDoc, req, res) => {
const tableId = req.params.tableId;
const rowIds = req.body;
const sandboxRes = await handleSandboxError(tableId, [], activeDoc.applyUserActions(
docSessionFromRequest(req),
[['BulkRemoveRecord', tableId, rowIds]]));
res.json(sandboxRes.retValues[0]);
}));
// Download full document
// TODO: look at download behavior if ActiveDoc is shutdown during call (cannot
// use withDoc wrapper)
this._app.get('/api/docs/:docId/download', canView, throttled(async (req, res) => {
// We want to be have a way download broken docs that ActiveDoc may not be able
// to load. So, if the user owns the document, we unconditionally let them
// download.
if (await this._isOwner(req)) {
try {
// We carefully avoid creating an ActiveDoc for the document being downloaded,
// in case it is broken in some way. It is convenient to be able to download
// broken files for diagnosis/recovery.
return await this._docWorker.downloadDoc(req, res, this._docManager.storageManager);
} catch (e) {
if (e.message && e.message.match(/does not exist yet/)) {
// The document has never been seen on file system / s3. It may be new, so
// we try again after having created an ActiveDoc for the document.
await this._getActiveDoc(req);
return this._docWorker.downloadDoc(req, res, this._docManager.storageManager);
} else {
throw e;
}
}
} else {
// If the user is not an owner, we load the document as an ActiveDoc, and then
// check if the user has download permissions.
const activeDoc = await this._getActiveDoc(req);
if (!await activeDoc.canDownload(docSessionFromRequest(req))) {
throw new Error('not authorized to download this document');
}
return this._docWorker.downloadDoc(req, res, this._docManager.storageManager);
}
}));
// Update records. The records to update are identified by their id column. Any invalid id fails
// the request and returns a 400 error code.
this._app.patch('/api/docs/:docId/tables/:tableId/data', canEdit, withDoc(async (activeDoc, req, res) => {
const tableId = req.params.tableId;
const columnValues = req.body;
const colNames = Object.keys(columnValues);
const rowIds = columnValues.id;
// sandbox expects no id column
delete columnValues.id;
await handleSandboxError(tableId, colNames, activeDoc.applyUserActions(
docSessionFromRequest(req),
[['BulkUpdateRecord', tableId, rowIds, columnValues]]));
res.json(null);
}));
// Reload a document forcibly (in fact this closes the doc, it will be automatically
// reopened on use).
this._app.post('/api/docs/:docId/force-reload', canEdit, throttled(async (req, res) => {
const activeDoc = await this._getActiveDoc(req);
await activeDoc.reloadDoc();
res.json(null);
}));
this._app.post('/api/docs/:docId/recover', canEdit, throttled(async (req, res) => {
const recoveryModeRaw = req.body.recoveryMode;
const recoveryMode = (typeof recoveryModeRaw === 'boolean') ? recoveryModeRaw : undefined;
if (!this._isOwner(req)) { throw new Error('Only owners can control recovery mode'); }
const activeDoc = await this._docManager.fetchDoc(docSessionFromRequest(req), getDocId(req), recoveryMode);
res.json({
recoveryMode: activeDoc.recoveryMode
});
}));
// DELETE /api/docs/:docId
// Delete the specified doc.
this._app.delete('/api/docs/:docId', canEditMaybeRemoved, throttled(async (req, res) => {
await this._removeDoc(req, res, true);
}));
// POST /api/docs/:docId/remove
// Soft-delete the specified doc. If query parameter "permanent" is set,
// delete permanently.
this._app.post('/api/docs/:docId/remove', canEditMaybeRemoved, throttled(async (req, res) => {
await this._removeDoc(req, res, isParameterOn(req.query.permanent));
}));
this._app.get('/api/docs/:docId/snapshots', canView, withDoc(async (activeDoc, req, res) => {
const {snapshots} = await activeDoc.getSnapshots();
res.json({snapshots});
}));
this._app.post('/api/docs/:docId/flush', canEdit, throttled(async (req, res) => {
const activeDocPromise = this._getActiveDocIfAvailable(req);
if (!activeDocPromise) {
// Only need to flush if doc is actually open.
res.json(false);
return;
}
const activeDoc = await activeDocPromise;
await activeDoc.flushDoc();
res.json(true);
}));
(core) support GRIST_WORKER_GROUP to place worker into an exclusive group Summary: In an emergency, we may want to serve certain documents with "old" workers as we fix problems. This diff adds some support for that. * Creates duplicate task definitions and services for staging and production doc workers (called grist-docs-staging2 and grist-docs-prod2), pulling from distinct docker tags (staging2 and prod2). The services are set to have zero workers until we need them. * These new workers are started with a new env variable `GRIST_WORKER_GROUP` set to `secondary`. * The `GRIST_WORKER_GROUP` variable, if set, makes the worker available to documents in the named group, and only that group. * An unauthenticated `/assign` endpoint is added to documents which, when POSTed to, checks that the doc is served by a worker in the desired group for that doc (as set manually in redis), and if not frees the doc up for reassignment. This makes it possible to move individual docs between workers without redeployments. The bash scripts added are a record of how the task definitions + services were created. The services could just have been copied manually, but the task definitions will need to be updated whenever the definitions for the main doc workers are updated, so it is worth scripting that. For example, if a certain document were to fail on a new deployment of Grist, but rolling back the full deployment wasn't practical: * Set prod2 tag in docker to desired codebase for that document * Set desired_count for grist-docs-prod2 service to non-zero * Set doc-<docid>-group for that doc in redis to secondary * Hit /api/docs/<docid>/assign to move the doc to grist-docs-prod2 (If the document needs to be reverted to a previous snapshot, that currently would need doing manually - could be made simpler, but not in scope of this diff). Test Plan: added tests Reviewers: dsagal Reviewed By: dsagal Differential Revision: https://phab.getgrist.com/D2649
4 years ago
// Administrative endpoint, that checks if a document is in the expected group,
// and frees it for reassignment if not. Has no effect if document is in the
// expected group. Does not require specific rights. Returns true if the document
// is freed up for reassignment, otherwise false.
this._app.post('/api/docs/:docId/assign', docExists, throttled(async (req, res) => {
const docId = getDocId(req);
const status = await this._docWorkerMap.getDocWorker(docId);
if (!status) { res.json(false); return; }
const workerGroup = await this._docWorkerMap.getWorkerGroup(status.docWorker.id);
const docGroup = await this._docWorkerMap.getDocGroup(docId);
if (docGroup === workerGroup) { res.json(false); return; }
const activeDoc = await this._getActiveDoc(req);
await activeDoc.flushDoc();
// flushDoc terminates once there's no pending operation on the document.
// There could still be async operations in progess. We mute their effect,
// as if they never happened.
activeDoc.docClients.interruptAllClients();
activeDoc.setMuted();
await activeDoc.shutdown();
await this._docWorkerMap.releaseAssignment(status.docWorker.id, docId);
res.json(true);
}));
// This endpoint cannot use withDoc since it is expected behavior for the ActiveDoc it
// starts with to become muted.
this._app.post('/api/docs/:docId/replace', canEdit, throttled(async (req, res) => {
const activeDoc = await this._getActiveDoc(req);
const options: DocReplacementOptions = {};
if (req.body.sourceDocId) {
options.sourceDocId = await this._confirmDocIdForRead(req, String(req.body.sourceDocId));
// We should make sure the source document has flushed recently.
// It may not be served by the same worker, so work through the api.
await fetch(this._grist.getHomeUrl(req, `/api/docs/${options.sourceDocId}/flush`), {
method: 'POST',
headers: {
...getTransitiveHeaders(req),
'Content-Type': 'application/json',
}
});
}
if (req.body.snapshotId) {
options.snapshotId = String(req.body.snapshotId);
}
await activeDoc.replace(options);
res.json(null);
}));
this._app.get('/api/docs/:docId/states', canView, withDoc(async (activeDoc, req, res) => {
const docSession = docSessionFromRequest(req);
res.json(await this._getStates(docSession, activeDoc));
}));
this._app.get('/api/docs/:docId/compare/:docId2', canView, withDoc(async (activeDoc, req, res) => {
const showDetails = isAffirmative(req.query.detail);
const docSession = docSessionFromRequest(req);
const {states} = await this._getStates(docSession, activeDoc);
const ref = await fetch(this._grist.getHomeUrl(req, `/api/docs/${req.params.docId2}/states`), {
headers: {
...getTransitiveHeaders(req),
'Content-Type': 'application/json',
}
});
const states2: DocState[] = (await ref.json()).states;
const left = states[0];
const right = states2[0];
if (!left || !right) {
// This should not arise unless there's a bug.
throw new Error('document with no history');
}
const rightHashes = new Set(states2.map(state => state.h));
const parent = states.find(state => rightHashes.has(state.h )) || null;
const leftChanged = parent && parent.h !== left.h;
const rightChanged = parent && parent.h !== right.h;
const summary = leftChanged ? (rightChanged ? 'both' : 'left') :
(rightChanged ? 'right' : (parent ? 'same' : 'unrelated'));
const comparison: DocStateComparison = {
left, right, parent, summary
};
if (showDetails && parent) {
// Calculate changes from the parent to the current version of this document.
const leftChanges = (await this._getChanges(docSession, activeDoc, states, parent.h,
'HEAD')).details!.rightChanges;
// Calculate changes from the (common) parent to the current version of the other document.
const url = `/api/docs/${req.params.docId2}/compare?left=${parent.h}`;
const rightChangesReq = await fetch(this._grist.getHomeUrl(req, url), {
headers: {
...getTransitiveHeaders(req),
'Content-Type': 'application/json',
}
});
const rightChanges = (await rightChangesReq.json()).details!.rightChanges;
// Add the left and right changes as details to the result.
comparison.details = { leftChanges, rightChanges };
}
res.json(comparison);
}));
// Give details about what changed between two versions of a document.
this._app.get('/api/docs/:docId/compare', canView, withDoc(async (activeDoc, req, res) => {
// This could be a relatively slow operation if actions are large.
const left = stringParam(req.query.left || 'HEAD');
const right = stringParam(req.query.right || 'HEAD');
const docSession = docSessionFromRequest(req);
const {states} = await this._getStates(docSession, activeDoc);
res.json(await this._getChanges(docSession, activeDoc, states, left, right));
}));
// Do an import targeted at a specific workspace. Although the URL fits ApiServer, this
// endpoint is handled only by DocWorker, so is handled here. (Note: this does not handle
// actual file uploads, so no worries here about large request bodies.)
this._app.post('/api/workspaces/:wid/import', expressWrap(async (req, res) => {
const userId = getUserId(req);
const wsId = integerParam(req.params.wid);
const uploadId = integerParam(req.body.uploadId);
const result = await this._docManager.importDocToWorkspace(userId, uploadId, wsId, req.body.browserSettings);
res.json(result);
}));
// Create a document. When an upload is included, it is imported as the initial
// state of the document. Otherwise a fresh empty document is created.
// A "timezone" option can be supplied.
// Documents are created "unsaved".
// TODO: support workspaceId option for creating regular documents, at which point
// existing import endpoint and doc creation endpoint can share implementation
// with this.
// Returns the id of the created document.
this._app.post('/api/docs', expressWrap(async (req, res) => {
const userId = getUserId(req);
let uploadId: number|undefined;
let parameters: {[key: string]: any};
if (req.is('multipart/form-data')) {
const formResult = await handleOptionalUpload(req, res);
if (formResult.upload) {
uploadId = formResult.upload.uploadId;
}
parameters = formResult.parameters || {};
} else {
parameters = req.body;
}
if (parameters.workspaceId) { throw new Error('workspaceId not supported'); }
const browserSettings: BrowserSettings = {};
if (parameters.timezone) { browserSettings.timezone = parameters.timezone; }
if (uploadId !== undefined) {
const result = await this._docManager.importDocToWorkspace(userId, uploadId, null,
browserSettings);
return res.json(result.id);
}
const isAnonymous = isAnonymousUser(req);
const {docId} = makeForkIds({userId, isAnonymous, trunkDocId: NEW_DOCUMENT_CODE,
trunkUrlId: NEW_DOCUMENT_CODE});
await this._docManager.fetchDoc(makeExceptionalDocSession('nascent', {
req: req as RequestWithLogin,
browserSettings
}), docId);
return res.status(200).json(docId);
}));
}
/**
* Check for read access to the given document, and return its
* canonical docId. Throws error if read access not available.
* This method is used for documents that are not the main document
* associated with the request, but are rather an extra source to be
* read from, so the access information is not cached in the
* request.
*/
private async _confirmDocIdForRead(req: Request, urlId: string): Promise<string> {
const userId = getUserId(req);
const org = (req as RequestWithLogin).org;
const docAuth = await makeDocAuthResult(this._dbManager.getDoc({urlId, userId, org}));
if (docAuth.error) { throw docAuth.error; }
assertAccess('viewers', docAuth);
return docAuth.docId!;
}
private _getActiveDoc(req: RequestWithLogin): Promise<ActiveDoc> {
return this._docManager.fetchDoc(docSessionFromRequest(req), getDocId(req));
}
private _getActiveDocIfAvailable(req: RequestWithLogin): Promise<ActiveDoc>|undefined {
return this._docManager.getActiveDoc(getDocId(req));
}
(core) support GRIST_WORKER_GROUP to place worker into an exclusive group Summary: In an emergency, we may want to serve certain documents with "old" workers as we fix problems. This diff adds some support for that. * Creates duplicate task definitions and services for staging and production doc workers (called grist-docs-staging2 and grist-docs-prod2), pulling from distinct docker tags (staging2 and prod2). The services are set to have zero workers until we need them. * These new workers are started with a new env variable `GRIST_WORKER_GROUP` set to `secondary`. * The `GRIST_WORKER_GROUP` variable, if set, makes the worker available to documents in the named group, and only that group. * An unauthenticated `/assign` endpoint is added to documents which, when POSTed to, checks that the doc is served by a worker in the desired group for that doc (as set manually in redis), and if not frees the doc up for reassignment. This makes it possible to move individual docs between workers without redeployments. The bash scripts added are a record of how the task definitions + services were created. The services could just have been copied manually, but the task definitions will need to be updated whenever the definitions for the main doc workers are updated, so it is worth scripting that. For example, if a certain document were to fail on a new deployment of Grist, but rolling back the full deployment wasn't practical: * Set prod2 tag in docker to desired codebase for that document * Set desired_count for grist-docs-prod2 service to non-zero * Set doc-<docid>-group for that doc in redis to secondary * Hit /api/docs/<docid>/assign to move the doc to grist-docs-prod2 (If the document needs to be reverted to a previous snapshot, that currently would need doing manually - could be made simpler, but not in scope of this diff). Test Plan: added tests Reviewers: dsagal Reviewed By: dsagal Differential Revision: https://phab.getgrist.com/D2649
4 years ago
private async _assertAccess(role: 'viewers'|'editors'|null, allowRemoved: boolean,
req: Request, res: Response, next: NextFunction) {
const scope = getDocScope(req);
allowRemoved = scope.showAll || scope.showRemoved || allowRemoved;
const docAuth = await getOrSetDocAuth(req as RequestWithLogin, this._dbManager, scope.urlId);
(core) support GRIST_WORKER_GROUP to place worker into an exclusive group Summary: In an emergency, we may want to serve certain documents with "old" workers as we fix problems. This diff adds some support for that. * Creates duplicate task definitions and services for staging and production doc workers (called grist-docs-staging2 and grist-docs-prod2), pulling from distinct docker tags (staging2 and prod2). The services are set to have zero workers until we need them. * These new workers are started with a new env variable `GRIST_WORKER_GROUP` set to `secondary`. * The `GRIST_WORKER_GROUP` variable, if set, makes the worker available to documents in the named group, and only that group. * An unauthenticated `/assign` endpoint is added to documents which, when POSTed to, checks that the doc is served by a worker in the desired group for that doc (as set manually in redis), and if not frees the doc up for reassignment. This makes it possible to move individual docs between workers without redeployments. The bash scripts added are a record of how the task definitions + services were created. The services could just have been copied manually, but the task definitions will need to be updated whenever the definitions for the main doc workers are updated, so it is worth scripting that. For example, if a certain document were to fail on a new deployment of Grist, but rolling back the full deployment wasn't practical: * Set prod2 tag in docker to desired codebase for that document * Set desired_count for grist-docs-prod2 service to non-zero * Set doc-<docid>-group for that doc in redis to secondary * Hit /api/docs/<docid>/assign to move the doc to grist-docs-prod2 (If the document needs to be reverted to a previous snapshot, that currently would need doing manually - could be made simpler, but not in scope of this diff). Test Plan: added tests Reviewers: dsagal Reviewed By: dsagal Differential Revision: https://phab.getgrist.com/D2649
4 years ago
if (role) { assertAccess(role, docAuth, {allowRemoved}); }
next();
}
/**
* Check if user is an owner of the document.
*/
private async _isOwner(req: Request) {
const scope = getDocScope(req);
const docAuth = await getOrSetDocAuth(req as RequestWithLogin, this._dbManager, scope.urlId);
return docAuth.access === 'owners';
}
// Helper to generate a 503 if the ActiveDoc has been muted.
private _checkForMute(activeDoc: ActiveDoc|undefined) {
if (activeDoc && activeDoc.muted) {
throw new ApiError('Document in flux - try again later', 503);
}
}
/**
* Throws an error if, during processing, the ActiveDoc becomes "muted". Also replaces any
* other error that may have occurred if the ActiveDoc becomes "muted", since the document
* shutting down during processing may have caused a variety of errors.
*
* Expects to be called within a handler that catches exceptions.
*/
private _requireActiveDoc(callback: WithDocHandler): RequestHandler {
return async (req, res) => {
let activeDoc: ActiveDoc|undefined;
try {
activeDoc = await this._getActiveDoc(req as RequestWithLogin);
await callback(activeDoc, req as RequestWithLogin, res);
if (!res.headersSent) { this._checkForMute(activeDoc); }
} catch (err) {
this._checkForMute(activeDoc);
throw err;
}
};
}
private async _getStates(docSession: OptDocSession, activeDoc: ActiveDoc): Promise<DocStates> {
const states = await activeDoc.getRecentStates(docSession);
return {
states,
};
}
/**
*
* Calculate changes between two document versions identified by leftHash and rightHash.
* If rightHash is the latest version of the document, the ActionSummary for it will
* contain a copy of updated and added rows.
*
* Currently will fail if leftHash is not an ancestor of rightHash (this restriction could
* be lifted, but is adequate for now).
*
*/
private async _getChanges(docSession: OptDocSession, activeDoc: ActiveDoc, states: DocState[],
leftHash: string, rightHash: string): Promise<DocStateComparison> {
const finder = new HashUtil(states);
const leftOffset = finder.hashToOffset(leftHash);
const rightOffset = finder.hashToOffset(rightHash);
if (rightOffset > leftOffset) {
throw new Error('Comparisons currently require left to be an ancestor of right');
}
const actionNums: number[] = states.slice(rightOffset, leftOffset).map(state => state.n);
const actions = (await activeDoc.getActions(actionNums)).reverse();
let totalAction = createEmptyActionSummary();
for (const action of actions) {
if (!action) { continue; }
const summary = summarizeAction(action);
totalAction = concatenateSummaries([totalAction, summary]);
}
const result: DocStateComparison = {
left: states[leftOffset],
right: states[rightOffset],
parent: states[leftOffset],
summary: (leftOffset === rightOffset) ? 'same' : 'right',
details: {
leftChanges: {tableRenames: [], tableDeltas: {}},
rightChanges: totalAction
}
};
return result;
}
private async _removeDoc(req: Request, res: Response, permanent: boolean) {
const scope = getDocScope(req);
const docId = getDocId(req);
if (permanent) {
const query = await this._dbManager.deleteDocument(scope);
this._dbManager.checkQueryResult(query); // fail immediately if deletion denied.
await this._docManager.deleteDoc(null, docId, true);
await sendReply(req, res, query);
} else {
await this._dbManager.softDeleteDocument(scope);
await sendOkReply(req, res);
}
await this._dbManager.flushSingleDocAuthCache(scope, docId);
await this._docManager.interruptDocClients(docId);
}
}
export function addDocApiRoutes(
(core) support GRIST_WORKER_GROUP to place worker into an exclusive group Summary: In an emergency, we may want to serve certain documents with "old" workers as we fix problems. This diff adds some support for that. * Creates duplicate task definitions and services for staging and production doc workers (called grist-docs-staging2 and grist-docs-prod2), pulling from distinct docker tags (staging2 and prod2). The services are set to have zero workers until we need them. * These new workers are started with a new env variable `GRIST_WORKER_GROUP` set to `secondary`. * The `GRIST_WORKER_GROUP` variable, if set, makes the worker available to documents in the named group, and only that group. * An unauthenticated `/assign` endpoint is added to documents which, when POSTed to, checks that the doc is served by a worker in the desired group for that doc (as set manually in redis), and if not frees the doc up for reassignment. This makes it possible to move individual docs between workers without redeployments. The bash scripts added are a record of how the task definitions + services were created. The services could just have been copied manually, but the task definitions will need to be updated whenever the definitions for the main doc workers are updated, so it is worth scripting that. For example, if a certain document were to fail on a new deployment of Grist, but rolling back the full deployment wasn't practical: * Set prod2 tag in docker to desired codebase for that document * Set desired_count for grist-docs-prod2 service to non-zero * Set doc-<docid>-group for that doc in redis to secondary * Hit /api/docs/<docid>/assign to move the doc to grist-docs-prod2 (If the document needs to be reverted to a previous snapshot, that currently would need doing manually - could be made simpler, but not in scope of this diff). Test Plan: added tests Reviewers: dsagal Reviewed By: dsagal Differential Revision: https://phab.getgrist.com/D2649
4 years ago
app: Application, docWorker: DocWorker, docWorkerMap: IDocWorkerMap, docManager: DocManager, dbManager: HomeDBManager,
grist: GristServer
) {
(core) support GRIST_WORKER_GROUP to place worker into an exclusive group Summary: In an emergency, we may want to serve certain documents with "old" workers as we fix problems. This diff adds some support for that. * Creates duplicate task definitions and services for staging and production doc workers (called grist-docs-staging2 and grist-docs-prod2), pulling from distinct docker tags (staging2 and prod2). The services are set to have zero workers until we need them. * These new workers are started with a new env variable `GRIST_WORKER_GROUP` set to `secondary`. * The `GRIST_WORKER_GROUP` variable, if set, makes the worker available to documents in the named group, and only that group. * An unauthenticated `/assign` endpoint is added to documents which, when POSTed to, checks that the doc is served by a worker in the desired group for that doc (as set manually in redis), and if not frees the doc up for reassignment. This makes it possible to move individual docs between workers without redeployments. The bash scripts added are a record of how the task definitions + services were created. The services could just have been copied manually, but the task definitions will need to be updated whenever the definitions for the main doc workers are updated, so it is worth scripting that. For example, if a certain document were to fail on a new deployment of Grist, but rolling back the full deployment wasn't practical: * Set prod2 tag in docker to desired codebase for that document * Set desired_count for grist-docs-prod2 service to non-zero * Set doc-<docid>-group for that doc in redis to secondary * Hit /api/docs/<docid>/assign to move the doc to grist-docs-prod2 (If the document needs to be reverted to a previous snapshot, that currently would need doing manually - could be made simpler, but not in scope of this diff). Test Plan: added tests Reviewers: dsagal Reviewed By: dsagal Differential Revision: https://phab.getgrist.com/D2649
4 years ago
const api = new DocWorkerApi(app, docWorker, docWorkerMap, docManager, dbManager, grist);
api.addEndpoints();
}
/**
* Catches the errors thrown by the sandbox, and converts to more descriptive ones (such as for
* invalid table names, columns, or rowIds) with better status codes. Accepts the table name, a
* list of column names in that table, and a promise for the result of the sandbox call.
*/
async function handleSandboxError<T>(tableId: string, colNames: string[], p: Promise<T>): Promise<T> {
try {
return await p;
} catch (e) {
if (e instanceof SandboxError) {
let match = e.message.match(/non-existent record #([0-9]+)/);
if (match) {
throw new ApiError(`Invalid row id ${match[1]}`, 400);
}
match = e.message.match(/\[Sandbox\] KeyError '(.*?)'/);
if (match) {
if (match[1] === tableId) {
throw new ApiError(`Table not found "${tableId}"`, 404);
} else if (colNames.includes(match[1])) {
throw new ApiError(`Invalid column "${match[1]}"`, 400);
}
}
throw new ApiError(`Error doing API call: ${e.message}`, 400);
}
throw e;
}
}
/**
* Options for returning results from a query about document data.
* Currently these option don't affect the query itself, only the
* results returned to the user.
*/
export interface QueryParameters {
sort?: string[]; // Columns to sort by (ascending order by default,
// prepend "-" for descending order).
limit?: number; // Limit on number of rows to return.
}
/**
* Extract a sort parameter from a request, if present. Follows
* https://jsonapi.org/format/#fetching-sorting for want of a better
* standard - comma separated, defaulting to ascending order, keys
* prefixed by "-" for descending order.
*
* The sort parameter can either be given as a query parameter, or
* as a header.
*/
function getSortParameter(req: Request): string[]|undefined {
const sortString: string|undefined = optStringParam(req.query.sort) || req.get('X-Sort');
if (!sortString) { return undefined; }
return sortString.split(',');
}
/**
* Extract a limit parameter from a request, if present. Should be a
* simple integer. The limit parameter can either be given as a query
* parameter, or as a header.
*/
function getLimitParameter(req: Request): number|undefined {
const limitString: string|undefined = optStringParam(req.query.limit) || req.get('X-Limit');
if (!limitString) { return undefined; }
const limit = parseInt(limitString, 10);
if (isNaN(limit)) { throw new Error('limit is not a number'); }
return limit;
}
/**
* Extract sort and limit parameters from request, if they are present.
*/
function getQueryParameters(req: Request): QueryParameters {
return {
sort: getSortParameter(req),
limit: getLimitParameter(req),
};
}
/**
* Sort table contents being returned. Sort keys with a '-' prefix
* are sorted in descending order, otherwise ascending. Contents are
* modified in place.
*/
function applySort(values: TableColValues, sort: string[]) {
if (!sort) { return values; }
const sortKeys = sort.map(key => key.replace(/^-/, ''));
const iteratees = sortKeys.map(key => {
if (!(key in values)) {
throw new Error(`unknown key ${key}`);
}
const col = values[key];
return (i: number) => col[i];
});
const sortSpec = sort.map((key, i) => (key.startsWith('-') ? -i - 1 : i + 1));
const index = values.id.map((_, i) => i);
const sortFunc = new SortFunc({
getColGetter(i) { return iteratees[i - 1]; },
getManualSortGetter() { return null; }
});
sortFunc.updateSpec(sortSpec);
index.sort(sortFunc.compare.bind(sortFunc));
for (const key of Object.keys(values)) {
const col = values[key];
values[key] = index.map(i => col[i]);
}
return values;
}
/**
* Truncate columns to the first N values. Columns are modified in place.
*/
function applyLimit(values: TableColValues, limit: number) {
// for no limit, or 0 limit, do not apply any restriction
if (!limit) { return values; }
for (const key of Object.keys(values)) {
values[key].splice(limit);
}
return values;
}
/**
* Apply query parameters to table contents. Contents are modified in place.
*/
export function applyQueryParameters(values: TableColValues, params: QueryParameters): TableColValues {
if (params.sort) { applySort(values, params.sort); }
if (params.limit) { applyLimit(values, params.limit); }
return values;
}