大型集合的 Firestore DeadlineExceeded 异常

Question

我正在尝试从 Google Firestore 读取更大的集合以进行测试和存档。当我尝试从包含超过 6k 个文档的集合中获取所有文档时，我遇到了一些有趣的错误。

朴素的 Python 解决方案

我的第一次尝试是使用 Python google-cloud-firestore（版本 0.30.0）库。

source_client = firestore.Client()
source = source_client.collection(collection)
source_data = source.get()

counter = 0
for f in source_data:
    app.logger.info(f.id)
    counter += 1
    if counter % 100 == 0:
        app.logger.info('%s %d', datetime.now(), counter)

    app.logger.info('%s Finally read all %d documents', datetime.now(), counter)

它给出以下输出：

INFO:flask.app:2018-11-08 09:49:03.923795 6400  
INFO:flask.app:2018-11-08 09:49:04.115410 6500  
... 
INFO:flask.app:2018-11-08 09:49:03.923795 6400
INFO:flask.app:2018-11-08 09:49:04.115410 6500
WARNING:flask.app:2018-11-08 09:49:04.128478 copy brocken by exception
Traceback (most recent call last):
  File "/home/carsten/projects/transfertool/venv/lib/python3.6/site-packages/flask/app.py", line 2309, in __call__
    return self.wsgi_app(environ, start_response)
  File "/home/carsten/projects/transfertool/venv/lib/python3.6/site-packages/flask/app.py", line 2295, in wsgi_app
    response = self.handle_exception(e)
  File "/home/carsten/projects/transfertool/venv/lib/python3.6/site-packages/flask/app.py", line 1741, in handle_exception
    reraise(exc_type, exc_value, tb)
  File "/home/carsten/projects/transfertool/venv/lib/python3.6/site-packages/flask/_compat.py", line 35, in reraise
    raise value
  File "/home/carsten/projects/transfertool/venv/lib/python3.6/site-packages/flask/app.py", line 2292, in wsgi_app
    response = self.full_dispatch_request()
  File "/home/carsten/projects/transfertool/venv/lib/python3.6/site-packages/flask/app.py", line 1815, in full_dispatch_request
    rv = self.handle_user_exception(e)
  File "/home/carsten/projects/transfertool/venv/lib/python3.6/site-packages/flask/app.py", line 1718, in handle_user_exception
    reraise(exc_type, exc_value, tb)
  File "/home/carsten/projects/transfertool/venv/lib/python3.6/site-packages/flask/_compat.py", line 35, in reraise
    raise value
  File "/home/carsten/projects/transfertool/venv/lib/python3.6/site-packages/flask/app.py", line 1813, in full_dispatch_request
    rv = self.dispatch_request()
  File "/home/carsten/projects/transfertool/venv/lib/python3.6/site-packages/flask/app.py", line 1799, in dispatch_request
    return self.view_functions[rule.endpoint](**req.view_args)
  File "/home/carsten/projects/transfertool/firestore/transfertool/main.py", line 142, in transfer
    count_collection(source_collection)
  File "/home/carsten/projects/transfertool/firestore/transfertool/main.py", line 94, in count_collection
    for f in source_collection.offset(1000).get():
  File "/home/carsten/projects/transfertool/venv/lib/python3.6/site-packages/google/cloud/firestore_v1beta1/query.py", line 588, in get
    for index, response_pb in enumerate(response_iterator):
  File "/home/carsten/projects/transfertool/venv/lib/python3.6/site-packages/google/api_core/grpc_helpers.py", line 83, in next
    six.raise_from(exceptions.from_grpc_error(exc), exc)
  File "<string>", line 3, in raise_from
    # Permission is hereby granted, free of charge, to any person obtaining a copy
google.api_core.exceptions.DeadlineExceeded: 504 Deadline Exceeded

这似乎是由配额引起的。即使我看不到它here。它似乎是基于时间的，因为当我在元素之间以小睡眠运行时，吞吐量会降低，并且在大约 50 秒后会出现异常。

使用 Python 进行分页

对于这个问题，这个库中有一个分页部分。由于我的应用程序不应该关心我尝试传输什么样的数据，我无法使用start_after 接口，但仍然有一个偏移接口，我至少可以使用它分批读取。

for f in source_collection.offset(last_read_offset).get():

只要last_read_offset 低于 1001，这会给我正确的结果。如果我从 1000 的偏移量开始，我可以得到结果，直到我从上面得到 google.api_core.exceptions.DeadlineExceeded exception。但是当我从更大的事情开始时，我会得到：

Traceback (most recent call last):
  File "/home/carsten/projects/transfertool/venv/lib/python3.6/site-packages/flask/app.py", line 2309, in __call__
    return self.wsgi_app(environ, start_response)
  File "/home/carsten/projects/transfertool/venv/lib/python3.6/site-packages/flask/app.py", line 2295, in wsgi_app
    response = self.handle_exception(e)
  File "/home/carsten/projects/transfertool/venv/lib/python3.6/site-packages/flask/app.py", line 1741, in handle_exception
    reraise(exc_type, exc_value, tb)
  File "/home/carsten/projects/transfertool/venv/lib/python3.6/site-packages/flask/_compat.py", line 35, in reraise
    raise value
  File "/home/carsten/projects/transfertool/venv/lib/python3.6/site-packages/flask/app.py", line 2292, in wsgi_app
    response = self.full_dispatch_request()
  File "/home/carsten/projects/transfertool/venv/lib/python3.6/site-packages/flask/app.py", line 1815, in full_dispatch_request
    rv = self.handle_user_exception(e)
  File "/home/carsten/projects/transfertool/venv/lib/python3.6/site-packages/flask/app.py", line 1718, in handle_user_exception
    reraise(exc_type, exc_value, tb)
  File "/home/carsten/projects/transfertool/venv/lib/python3.6/site-packages/flask/_compat.py", line 35, in reraise
    raise value
  File "/home/carsten/projects/transfertool/venv/lib/python3.6/site-packages/flask/app.py", line 1813, in full_dispatch_request
    rv = self.dispatch_request()
  File "/home/carsten/projects/transfertool/venv/lib/python3.6/site-packages/flask/app.py", line 1799, in dispatch_request
    return self.view_functions[rule.endpoint](**req.view_args)
  File "/home/carsten/projects/transfertool/firestore/transfertool/main.py", line 144, in transfer
    count_collection(source_collection)
  File "/home/carsten/projects/transfertool/firestore/transfertool/main.py", line 94, in count_collection
    for f in source_collection.offset(1001).get():
  File "/home/carsten/projects/transfertool/venv/lib/python3.6/site-packages/google/cloud/firestore_v1beta1/query.py", line 599, in get
    raise ValueError(msg)
ValueError: Unexpected server response. All responses other than the first must contain a document. The response at index 1 was
read_time {
  seconds: 1541668338
  nanos: 420813000
}
skipped_results: 1

查看库代码，后端似乎正在发送一条被解释为无效的消息。

通过 node.js 重试

好吧，也许我的代码或 Python 客户端库有问题。让我们尝试使用节点。

const admin = require('firebase-admin');
admin.initializeApp({
    credential: admin.credential.applicationDefault()
});

var db = admin.firestore();
admin.firestore().settings( { timestampsInSnapshots: true })
var counter = 0

console.log('Read collection')
db.collection(collection).get()
    .then(querySnapshot => {
        querySnapshot.forEach(documentSnapshot => {
            counter++;
        });
        console.log(counter)
    })
    .catch( error => {
        console.log(error)
});

这与 python 库的作用相同，即使超时更明显是 60 秒。

[2018-11-09T08:36:30.992Z] App listening on port 8080
[2018-11-09T08:36:30.993Z] Press Ctrl+C to quit.
[2018-11-09T08:36:37.390Z] Read collection
[2018-11-09T08:37:37.406Z] { Error: 4 DEADLINE_EXCEEDED: Deadline Exceeded
    at Object.exports.createStatusError (/home/carsten/projects/node_modules/grpc/src/common.js:87:15)
    at ClientReadableStream._emitStatusIfDone (/home/carsten/projects/node_modules/grpc/src/client.js:235:26)
    at ClientReadableStream._readsDone (/home/carsten/projects/node_modules/grpc/src/client.js:201:8)
    at /home/carsten/projects/node_modules/grpc/src/client_interceptors.js:679:15
  code: 4,
  metadata: Metadata { _internal_repr: {} },
  details: 'Deadline Exceeded' }

有没有人有类似的经历或很好的提示如何继续？

PS：exportDocument/importDocument 接口不够用，有时我们需要在读取后调整数据。而且我不知道 Firestore 将哪种格式导出到 Google Cloud Storage 或如何转换它。

编辑：golang

为了它，我尝试了 golang api。

log.Println("Collecting data")
snapshotIter := client.Collection(collection.(string)).Documents(ctx)
defer snapshotIter.Stop()

if err != nil {
    log.Fatalln(err)
}

i := 0

for {
    _, err := snapshotIter.Next()

    if err == iterator.Done {
        break
    }
    if err != nil {
        log.Fatalln(err)
    }

    if i % 100 == 0{
        log.Println(i)
    }
    i++
}

log.Println("Done")

这会遇到与预期相同的超时。

2018/11/12 15:01:20 Collecting data
2018/11/12 15:01:21 0
2018/11/12 15:01:21 100
2018/11/12 15:01:21 200
2018/11/12 15:01:21 300
2018/11/12 15:01:21 400
2018/11/12 15:01:22 500
2018/11/12 15:01:22 600
2018/11/12 15:01:22 700
....
2018/11/12 15:02:22 29800
2018/11/12 15:02:23 29900
2018/11/12 15:02:23 rpc error: code = DeadlineExceeded desc = The datastore operation timed out, or the data was temporarily unavailable.

但除此之外，偏移量也可以正常工作：

snapshotIter := client.Collection(collection.(string)).Offset(30000).Documents(ctx)

Answer 1

在我的情况下，我在获取整个集合时遇到了这个错误。它甚至不是那么大的集合，但我猜集合中的文档很大。我做了一个分页更新。这是一个节点 firebase 函数：

let lastReadDoc = false;


let lastDoc: string = '';
  const limitRecordCount = 10;
  do {
    await db
      .collection('something/' + somethingId + '/somethingcollection')
      .orderBy('id')
      .limit(limitRecordCount)
      .startAfter(lastDoc)
      .get()
      .then((snapshot: any) => {
        let counter = 0;
        snapshot.docs.forEach((doc: any) => {
          const docData = doc.data();
          if (lastDoc !== docData.id) {
            lastDoc = docData.id;
            counter = counter + 1;
          }
          // ***********************
          // business logic per doc here
          // ***********************
        });
        if (counter < limitRecordCount) {
          lastReadDoc = true;
        }
      })
      .catch((err: any) => {
        lastReadDoc = true;
        console.log('Error getting booking documents', err);
      });
  } while (lastReadDoc === false);

Answer 2

在 firebase 支持团队的帮助下，我们发现 python 客户端 api 确实存在错误。在下一个版本中会有一个错误修复。很可能它将使 python 库能够按 documentid 排序，因此使用start_after()。

在此之前，您有两种可能的解决方案：

1. 使用另一个字段进行排序并使用start_after()

2. 使用带有分页功能的 node.js 库，例如：

var db = admin.firestore();
admin.firestore().settings({ timestampsInSnapshots: true });
function readNextPage(lastReadDoc) {
  let query = db
    .collection(collection)
    .orderBy(admin.firestore.FieldPath.documentId())
    .limit(100);
}