1
0
Fork 0
mirror of https://we.phorge.it/source/phorge.git synced 2024-12-29 17:00:59 +01:00
phorge-phorge/support/aphlict/server/lib/AphlictClientServer.js

191 lines
4.8 KiB
JavaScript
Raw Normal View History

'use strict';
var JX = require('./javelin').JX;
require('./AphlictListenerList');
require('./AphlictLog');
Namespace Aphlict clients by request path, plus other fixes Summary: Fixes T7130. Fixes T7041. Fixes T7012. Major change here is partitioning clients. In the Phacility cluster, being able to get a huge pile of instances on a single server -- without needing to run a process per instance -- is desirable. To accomplish this, just bucket clients by the path they connect with. This will let us set client URIs to `/instancename/` and then route connections to a small set of servers. This degrades cleanly in the common case and has no effect on installs which don't do instancing. Also fix two unrelated issues: - Fix the timeouts, which were incorrectly initializing in `open()` (which is called during reconnect, causing them to reset every time). Instead, initialize in the constructor. Cap timeout at 5 minutes. - Probably fix subscriptions, which were using a property with an object definition. Since this is by-ref, all concrete instances of the object share the same property, so all users would be subscribed to everything. Probably. Test Plan: - Hit notification status page, saw version bump and instance/path name. - Saw instance/path name in client and server logs. - Stopped server, saw reconnects after 2, 4, 16, ... seconds. - Sent test notification; received test notification. - Didn't explicitly test the subscription thing but it should be obvious by looking at `/notification/status/` shortly after a push. Reviewers: joshuaspence, btrahan Reviewed By: btrahan Subscribers: epriestley Maniphest Tasks: T7041, T7012, T7130 Differential Revision: https://secure.phabricator.com/D11769
2015-02-16 20:31:15 +01:00
var url = require('url');
var util = require('util');
var WebSocket = require('ws');
JX.install('AphlictClientServer', {
construct: function(server) {
server.on('request', JX.bind(this, this._onrequest));
this._server = server;
Namespace Aphlict clients by request path, plus other fixes Summary: Fixes T7130. Fixes T7041. Fixes T7012. Major change here is partitioning clients. In the Phacility cluster, being able to get a huge pile of instances on a single server -- without needing to run a process per instance -- is desirable. To accomplish this, just bucket clients by the path they connect with. This will let us set client URIs to `/instancename/` and then route connections to a small set of servers. This degrades cleanly in the common case and has no effect on installs which don't do instancing. Also fix two unrelated issues: - Fix the timeouts, which were incorrectly initializing in `open()` (which is called during reconnect, causing them to reset every time). Instead, initialize in the constructor. Cap timeout at 5 minutes. - Probably fix subscriptions, which were using a property with an object definition. Since this is by-ref, all concrete instances of the object share the same property, so all users would be subscribed to everything. Probably. Test Plan: - Hit notification status page, saw version bump and instance/path name. - Saw instance/path name in client and server logs. - Stopped server, saw reconnects after 2, 4, 16, ... seconds. - Sent test notification; received test notification. - Didn't explicitly test the subscription thing but it should be obvious by looking at `/notification/status/` shortly after a push. Reviewers: joshuaspence, btrahan Reviewed By: btrahan Subscribers: epriestley Maniphest Tasks: T7041, T7012, T7130 Differential Revision: https://secure.phabricator.com/D11769
2015-02-16 20:31:15 +01:00
this._lists = {};
this._adminServers = [];
},
properties: {
logger: null,
adminServers: null
},
members: {
_server: null,
Namespace Aphlict clients by request path, plus other fixes Summary: Fixes T7130. Fixes T7041. Fixes T7012. Major change here is partitioning clients. In the Phacility cluster, being able to get a huge pile of instances on a single server -- without needing to run a process per instance -- is desirable. To accomplish this, just bucket clients by the path they connect with. This will let us set client URIs to `/instancename/` and then route connections to a small set of servers. This degrades cleanly in the common case and has no effect on installs which don't do instancing. Also fix two unrelated issues: - Fix the timeouts, which were incorrectly initializing in `open()` (which is called during reconnect, causing them to reset every time). Instead, initialize in the constructor. Cap timeout at 5 minutes. - Probably fix subscriptions, which were using a property with an object definition. Since this is by-ref, all concrete instances of the object share the same property, so all users would be subscribed to everything. Probably. Test Plan: - Hit notification status page, saw version bump and instance/path name. - Saw instance/path name in client and server logs. - Stopped server, saw reconnects after 2, 4, 16, ... seconds. - Sent test notification; received test notification. - Didn't explicitly test the subscription thing but it should be obvious by looking at `/notification/status/` shortly after a push. Reviewers: joshuaspence, btrahan Reviewed By: btrahan Subscribers: epriestley Maniphest Tasks: T7041, T7012, T7130 Differential Revision: https://secure.phabricator.com/D11769
2015-02-16 20:31:15 +01:00
_lists: null,
getListenerList: function(instance) {
if (!this._lists[instance]) {
this._lists[instance] = new JX.AphlictListenerList(instance);
Namespace Aphlict clients by request path, plus other fixes Summary: Fixes T7130. Fixes T7041. Fixes T7012. Major change here is partitioning clients. In the Phacility cluster, being able to get a huge pile of instances on a single server -- without needing to run a process per instance -- is desirable. To accomplish this, just bucket clients by the path they connect with. This will let us set client URIs to `/instancename/` and then route connections to a small set of servers. This degrades cleanly in the common case and has no effect on installs which don't do instancing. Also fix two unrelated issues: - Fix the timeouts, which were incorrectly initializing in `open()` (which is called during reconnect, causing them to reset every time). Instead, initialize in the constructor. Cap timeout at 5 minutes. - Probably fix subscriptions, which were using a property with an object definition. Since this is by-ref, all concrete instances of the object share the same property, so all users would be subscribed to everything. Probably. Test Plan: - Hit notification status page, saw version bump and instance/path name. - Saw instance/path name in client and server logs. - Stopped server, saw reconnects after 2, 4, 16, ... seconds. - Sent test notification; received test notification. - Didn't explicitly test the subscription thing but it should be obvious by looking at `/notification/status/` shortly after a push. Reviewers: joshuaspence, btrahan Reviewed By: btrahan Subscribers: epriestley Maniphest Tasks: T7041, T7012, T7130 Differential Revision: https://secure.phabricator.com/D11769
2015-02-16 20:31:15 +01:00
}
return this._lists[instance];
Namespace Aphlict clients by request path, plus other fixes Summary: Fixes T7130. Fixes T7041. Fixes T7012. Major change here is partitioning clients. In the Phacility cluster, being able to get a huge pile of instances on a single server -- without needing to run a process per instance -- is desirable. To accomplish this, just bucket clients by the path they connect with. This will let us set client URIs to `/instancename/` and then route connections to a small set of servers. This degrades cleanly in the common case and has no effect on installs which don't do instancing. Also fix two unrelated issues: - Fix the timeouts, which were incorrectly initializing in `open()` (which is called during reconnect, causing them to reset every time). Instead, initialize in the constructor. Cap timeout at 5 minutes. - Probably fix subscriptions, which were using a property with an object definition. Since this is by-ref, all concrete instances of the object share the same property, so all users would be subscribed to everything. Probably. Test Plan: - Hit notification status page, saw version bump and instance/path name. - Saw instance/path name in client and server logs. - Stopped server, saw reconnects after 2, 4, 16, ... seconds. - Sent test notification; received test notification. - Didn't explicitly test the subscription thing but it should be obvious by looking at `/notification/status/` shortly after a push. Reviewers: joshuaspence, btrahan Reviewed By: btrahan Subscribers: epriestley Maniphest Tasks: T7041, T7012, T7130 Differential Revision: https://secure.phabricator.com/D11769
2015-02-16 20:31:15 +01:00
},
getHistory: function(age) {
var results = [];
var servers = this.getAdminServers();
for (var ii = 0; ii < servers.length; ii++) {
var messages = servers[ii].getHistory(age);
for (var jj = 0; jj < messages.length; jj++) {
results.push(messages[jj]);
}
}
return results;
},
log: function() {
var logger = this.getLogger();
if (!logger) {
return;
}
logger.log.apply(logger, arguments);
return this;
},
_onrequest: function(request, response) {
// The websocket code upgrades connections before they get here, so
// this only handles normal HTTP connections. We just fail them with
// a 501 response.
response.writeHead(501);
response.end('HTTP/501 Use Websockets\n');
},
_parseInstanceFromPath: function(path) {
// If there's no "~" marker in the path, it's not an instance name.
// Users sometimes configure nginx or Apache to proxy based on the
// path.
if (path.indexOf('~') === -1) {
return 'default';
}
var instance = path.split('~')[1];
// Remove any "/" characters.
instance = instance.replace(/\//g, '');
if (!instance.length) {
return 'default';
}
return instance;
},
listen: function() {
var self = this;
var server = this._server.listen.apply(this._server, arguments);
var wss = new WebSocket.Server({server: server});
// This function checks for upgradeReq which is only available in
// ws2 by default, not ws3. See T12755 for more information.
wss.on('connection', function(ws, request) {
if ('upgradeReq' in ws) {
request = ws.upgradeReq;
}
var path = url.parse(request.url).pathname;
var instance = self._parseInstanceFromPath(path);
var listener = self.getListenerList(instance).addListener(ws);
function log() {
self.log(
util.format('<%s>', listener.getDescription()) +
' ' +
util.format.apply(null, arguments));
}
log('Connected from %s.', ws._socket.remoteAddress);
ws.on('message', function(data) {
log('Received message: %s', data);
var message;
try {
message = JSON.parse(data);
} catch (err) {
log('Message is invalid: %s', err.message);
return;
}
switch (message.command) {
case 'subscribe':
log(
'Subscribed to: %s',
JSON.stringify(message.data));
listener.subscribe(message.data);
break;
case 'unsubscribe':
log(
'Unsubscribed from: %s',
JSON.stringify(message.data));
listener.unsubscribe(message.data);
break;
case 'replay':
var age = message.data.age || 60000;
var min_age = (new Date().getTime() - age);
var old_messages = self.getHistory(min_age);
for (var ii = 0; ii < old_messages.length; ii++) {
var old_message = old_messages[ii];
if (!listener.isSubscribedToAny(old_message.subscribers)) {
continue;
}
try {
listener.writeMessage(old_message);
} catch (error) {
break;
}
}
break;
case 'ping':
var pong = {
type: 'pong'
};
try {
listener.writeMessage(pong);
} catch (error) {
// Ignore any issues here, we'll clean up elsewhere.
}
break;
default:
log(
'Unrecognized command "%s".',
message.command || '<undefined>');
}
});
ws.on('close', function() {
self.getListenerList(instance).removeListener(listener);
log('Disconnected.');
});
});
}
}
});