Google Cloud Pubsub Data lost

Question

I'm experiencing a problem with GCP pubsub where a small percentage of data was lost when publishing thousands of messages in couple seconds.

I'm logging both message_id from pubsub and a session_id unique to each message on both the publishing end as well as the receiving end, and the result I'm seeing is that some message on the receiving end has same session_id, but different message_id. Also, some messages were missing.

For example, in one test I send 5,000 messages to pubsub, and exactly 5,000 messages were received, with 8 messages lost. The log lost messages look like this:

MISSING sessionId:sessionId: 731 (missing in log from pull request, but present in log from Flask API)

messageId FOUND: messageId:108562396466545

API: 200 **** sessionId: 731, messageId:108562396466545 ******(Log from Flask API)

Pubsub: sessionId: 730, messageId:108562396466545(Log from pull request)

And the duplicates looks like:

======= Duplicates FOUND on sessionId: 730=======

sessionId: 730, messageId:108562396466545

sessionId: 730, messageId:108561339282318

(both are logs from pull request)

All missing data and duplicates look like this.

From the above example, it is clear that some messages has taken the message_id of another message, and has been sent twice with two different message_ids.

I wonder if anyone would help me figure out what is going on? Thanks in advance.

Code

I have an API sending message to pubsub, which looks like this:

from flask import Flask, request, jsonify, render_template
from flask_cors import CORS, cross_origin
import simplejson as json
from google.cloud import pubsub
from functools import wraps
import re
import json


app = Flask(__name__)
ps = pubsub.Client()

...

@app.route('/publish', methods=['POST'])
@cross_origin()
@json_validator
def publish_test_topic():
    pubsub_topic = 'test_topic'
    data = request.data

    topic = ps.topic(pubsub_topic)

    event = json.loads(data)

    messageId = topic.publish(data)
    return '200 **** sessionId: ' + str(event["sessionId"]) + ", messageId:" + messageId + " ******"

And this is the code I used to read from pubsub:

from google.cloud import pubsub import re import json

ps = pubsub.Client()
topic = ps.topic('test-xiu')
sub = topic.subscription('TEST-xiu')

max_messages = 1
stop = False

messages = []

class Message(object):
    """docstring for Message."""
    def __init__(self, sessionId, messageId):
        super(Message, self).__init__()
        self.seesionId = sessionId
        self.messageId = messageId


def pull_all():
    while stop == False:

        m = sub.pull(max_messages = max_messages, return_immediately = False)

        for data in m:
            ack_id = data[0]
            message = data[1]
            messageId = message.message_id
            data = message.data
            event = json.loads(data)
            sessionId = str(event["sessionId"])
            messages.append(Message(sessionId = sessionId, messageId = messageId))

            print '200 **** sessionId: ' + sessionId + ", messageId:" + messageId + " ******"

            sub.acknowledge(ack_ids = [ack_id])

pull_all()

For generating session_id, sending request & logging response from API:

// generate trackable sessionId
var sessionId = 0

var increment_session_id = function () {
  sessionId++;
  return sessionId;
}

var generate_data = function () {
  var data = {};
  // data.sessionId = faker.random.uuid();
  data.sessionId = increment_session_id();
  data.user = get_rand(userList);
  data.device = get_rand(deviceList);
  data.visitTime = new Date;
  data.location = get_rand(locationList);
  data.content = get_rand(contentList);

  return data;
}

var sendData = function (url, payload) {
  var request = $.ajax({
    url: url,
    contentType: 'application/json',
    method: 'POST',
    data: JSON.stringify(payload),
    error: function (xhr, status, errorThrown) {
      console.log(xhr, status, errorThrown);
      $('.result').prepend("<pre id='json'>" + JSON.stringify(xhr, null, 2) + "</pre>")
      $('.result').prepend("<div>errorThrown: " + errorThrown + "</div>")
      $('.result').prepend("<div>======FAIL=======</div><div>status: " + status + "</div>")
    }
  }).done(function (xhr) {
    console.log(xhr);
    $('.result').prepend("<div>======SUCCESS=======</div><pre id='json'>" + JSON.stringify(payload, null, 2) + "</pre>")
  })
}

$(submit_button).click(function () {
  var request_num = get_request_num();
  var request_url = get_url();
  for (var i = 0; i < request_num; i++) {
    var data = generate_data();
    var loadData = changeVerb(data, 'load');
    sendData(request_url, loadData);
  }
}) 

UPDATE

I made a change on the API, and the issue seems to go away. The changes I made was instead of using one pubsub.Client() for all request, I initialized a client for every single request coming in. The new API looks like:

from flask import Flask, request, jsonify, render_template
from flask_cors import CORS, cross_origin
import simplejson as json
from google.cloud import pubsub
from functools import wraps
import re
import json


app = Flask(__name__)

...

@app.route('/publish', methods=['POST'])
@cross_origin()
@json_validator
def publish_test_topic():

    ps = pubsub.Client()


    pubsub_topic = 'test_topic'
    data = request.data

    topic = ps.topic(pubsub_topic)

    event = json.loads(data)

    messageId = topic.publish(data)
    return '200 **** sessionId: ' + str(event["sessionId"]) + ", messageId:" + messageId + " ******"

Answer 1

Talked with some guy from Google, and it seems to be an issue with the Python Client:

The consensus on our side is that there is a thread-safety problem in the current python client. The client library is being rewritten almost from scratch as we speak, so I don't want to pursue any fixes in the current version. We expect the new version to become available by end of June.

Running the current code with thread_safe: false in app.yaml or better yet just instantiating the client in every call should is the work around -- the solution you found.

For detailed solution, please see the Update in the question

Answer 2

Google Cloud Pub/Sub message IDs are unique. It should not be possible for "some messages [to] taken the message_id of another message." The fact that message ID 108562396466545 was seemingly received means that Pub/Sub did deliver the message to the subscriber and was not lost.

I recommend you check how your session_ids are generated to ensure that they are indeed unique and that there is exactly one per message. Searching for the sessionId in your JSON via a regular expression search seems a little strange. You would be better off parsing this JSON into an actual object and accessing fields that way.

In general, duplicate messages in Cloud Pub/Sub are always possible; the system guarantees at-least-once delivery. Those messages can be delivered with the same message ID if the duplication happens on the subscribe side (e.g., the ack is not processed in time) or with a different message ID (e.g., if the publish of the message is retried after an error like a deadline exceeded).