Question

我有一个API端点，它只执行读取请求。通常，整个操作可以在100ms下完成

[03/Aug/2015:19:35:53 -0700] "GET /query?email=xxx%40aol.com&hash=xxx HTTP/1.1" 200 186 - "myapp-1.0.6o" "xxx.appspot.com" ms=74 cpu_ms=27 cpm_usd=0.000021 instance=00c61b117c55f2b00cdd73904665675ced040765 app_engine_release=1.9.24

但是，我可以看到一些请求并且不常见，最多可以激活20秒！

[03/Aug/2015:18:59:50 -0700] "GET /query?email=yyy%40gmail.com&hash=yyy HTTP/1.1" 200 193 - "myapp-1.0.6o" "xxx.appspot.com" ms=18288 cpu_ms=61 cpm_usd=0.000022 instance=00c61b117c55f2b00cdd73904665675ced040765 app_engine_release=1.9.24

现在，服务器负载仍然很轻。它会在一分钟内收到1-5个请求。我们仍然处于免费配额使用状态。

API python代码非常简单。我认为这不是这种缓慢行动的罪魁祸首。

class User(ndb.Model):
    email = ndb.StringProperty(required = True)
    timestamp = ndb.DateTimeProperty(required = True)

class QueryHandler(webapp2.RequestHandler):
    def get(self):
        email = self.request.get('email')
        hash = self.request.get('hash')
        expected_hash = Utils.hash(email)

        result = {
            'email' : email,
            'user_timestamp' : 0,
            'server_timestamp' : 0,
            'free_trial_duration' : _FREE_TRIAL_DURATION
        }

        user_timestamp = 0

        if hash == expected_hash:
            user_timestamp = memcache.get(email)

            if user_timestamp is None:                
                user = User.get_by_id(email)
                if user is not None:
                    user_timestamp = int(time.mktime(user.timestamp.timetuple()))
                    result['user_timestamp'] = user_timestamp
                    memcache.add(email, user_timestamp, _MEMCACHE_DURATION)
            else:
                result['user_timestamp'] = user_timestamp

        else:
            logging.debug('QueryHandler, email = ' + email + ', hash = ' + hash + ', expected_hash = ' + expected_hash)

        server_timestamp = max(int(time.time()), user_timestamp)
        result['server_timestamp'] = server_timestamp
        self.response.headers['Content-Type'] = 'application/json'
        json_result = json.encode(result)
        self.response.out.write(json_result)
        logging.debug('QueryHandler, result = ' + json_result)


app = webapp2.WSGIApplication([
    ('/query', QueryHandler),
], debug=False)

Google App Engine服务器质量似乎更多。有什么我们可以做的，以避免API消费的不合理缓慢？

Google App Engine的性能不一致

0 个答案: