解決Scrapy-Redis爬取完畢之後繼續空跑的問題
阿新 • • 發佈:2018-12-03
1. 背景
根據scrapy-redis分散式爬蟲的原理,多臺爬蟲主機共享一個爬取佇列。當爬取佇列中存在request時,爬蟲就會取出request進行爬取,如果爬取佇列中不存在request時,爬蟲就會處於等待狀態,行如下:
E:\Miniconda\python.exe E:/PyCharmCode/redisClawerSlaver/redisClawerSlaver/spiders/main.py 2017-12-12 15:54:18 [scrapy.utils.log] INFO: Scrapy 1.4.0 started (bot: scrapybot) 2017-12-12 15:54:18 [scrapy.utils.log] INFO: Overridden settings: {'SPIDER_LOADER_WARN_ONLY': True} 2017-12-12 15:54:18 [scrapy.middleware] INFO: Enabled extensions: ['scrapy.extensions.corestats.CoreStats', 'scrapy.extensions.telnet.TelnetConsole', 'scrapy.extensions.logstats.LogStats'] 2017-12-12 15:54:18 [myspider_redis] INFO: Reading start URLs from redis key 'myspider:start_urls' (batch size: 110, encoding: utf-8 2017-12-12 15:54:18 [scrapy.middleware] INFO: Enabled downloader middlewares: ['scrapy.downloadermiddlewares.httpauth.HttpAuthMiddleware', 'scrapy.downloadermiddlewares.downloadtimeout.DownloadTimeoutMiddleware', 'scrapy.downloadermiddlewares.defaultheaders.DefaultHeadersMiddleware', 'redisClawerSlaver.middlewares.ProxiesMiddleware', 'redisClawerSlaver.middlewares.HeadersMiddleware', 'scrapy.downloadermiddlewares.retry.RetryMiddleware', 'scrapy.downloadermiddlewares.redirect.MetaRefreshMiddleware', 'scrapy.downloadermiddlewares.httpcompression.HttpCompressionMiddleware', 'scrapy.downloadermiddlewares.redirect.RedirectMiddleware', 'scrapy.downloadermiddlewares.httpproxy.HttpProxyMiddleware', 'scrapy.downloadermiddlewares.stats.DownloaderStats'] 2017-12-12 15:54:18 [scrapy.middleware] INFO: Enabled spider middlewares: ['scrapy.spidermiddlewares.httperror.HttpErrorMiddleware', 'scrapy.spidermiddlewares.offsite.OffsiteMiddleware', 'scrapy.spidermiddlewares.referer.RefererMiddleware', 'scrapy.spidermiddlewares.urllength.UrlLengthMiddleware', 'scrapy.spidermiddlewares.depth.DepthMiddleware'] 2017-12-12 15:54:18 [scrapy.middleware] INFO: Enabled item pipelines: ['redisClawerSlaver.pipelines.ExamplePipeline', 'scrapy_redis.pipelines.RedisPipeline'] 2017-12-12 15:54:18 [scrapy.core.engine] INFO: Spider opened 2017-12-12 15:54:18 [scrapy.extensions.logstats] INFO: Crawled 0 pages (at 0 pages/min), scraped 0 items (at 0 items/min) 2017-12-12 15:55:18 [scrapy.extensions.logstats] INFO: Crawled 0 pages (at 0 pages/min), scraped 0 items (at 0 items/min) 2017-12-12 15:56:18 [scrapy.extensions.logstats] INFO: Crawled 0 pages (at 0 pages/min), scraped 0 items (at 0 items/min)
- 可是,如果所有的request都已經爬取完畢了呢?這件事爬蟲程式是不知道的,它無法區分結束和空窗期狀態的不同,所以會一直處於上面的那種等待狀態,也就是我們說的空跑。
- 那有沒有辦法讓爬蟲區分這種情況,自動結束呢?
2. 環境
- 系統:win7
- scrapy-redis
- redis 3.0.5
- python 3.6.1
3. 解決方案
- 從背景介紹來看,基於scrapy-redis分散式爬蟲的原理,爬蟲結束是一個很模糊的概念,在爬蟲爬取過程中,爬取佇列是一個不斷動態變化的過程,隨著request的爬取,又會有新的request進入爬取佇列。進進出出。爬取速度高於填充速度,就會有佇列空窗期(爬取佇列中,某一段時間會出現沒有request的情況),爬取速度低於填充速度,就不會出現空窗期。
- 所以,下面這兩種方案都是一種大概的思路。
3.1. 利用scrapy的關閉spider擴充套件功能
# 關閉spider擴充套件 class scrapy.contrib.closespider.CloseSpider 當某些狀況發生,spider會自動關閉。每種情況使用指定的關閉原因。 關閉spider的情況可以通過下面的設定項配置: CLOSESPIDER_TIMEOUT CLOSESPIDER_ITEMCOUNT CLOSESPIDER_PAGECOUNT CLOSESPIDER_ERRORCOUNT
- CLOSESPIDER_TIMEOUT
CLOSESPIDER_TIMEOUT
預設值: 0
一個整數值,單位為秒。如果一個spider在指定的秒數後仍在執行, 它將以 closespider_timeout 的原因被自動關閉。 如果值設定為0(或者沒有設定),spiders不會因為超時而關閉。
- CLOSESPIDER_ITEMCOUNT
CLOSESPIDER_ITEMCOUNT
預設值: 0
一個整數值,指定條目的個數。如果spider爬取條目數超過了指定的數, 並且這些條目通過item pipeline傳遞,spider將會以 closespider_itemcount 的原因被自動關閉。
- CLOSESPIDER_PAGECOUNT
CLOSESPIDER_PAGECOUNT
0.11 新版功能.
預設值: 0
一個整數值,指定最大的抓取響應(reponses)數。 如果spider抓取數超過指定的值,則會以 closespider_pagecount 的原因自動關閉。 如果設定為0(或者未設定),spiders不會因為抓取的響應數而關閉。
- CLOSESPIDER_ERRORCOUNT
CLOSESPIDER_ERRORCOUNT
0.11 新版功能.
預設值: 0
一個整數值,指定spider可以接受的最大錯誤數。 如果spider生成多於該數目的錯誤,它將以 closespider_errorcount 的原因關閉。 如果設定為0(或者未設定),spiders不會因為發生錯誤過多而關閉。
- 示例:開啟 settings.py,新增一個配置項,如下
# 爬蟲執行超過23.5小時,如果爬蟲還沒有結束,則自動關閉
CLOSESPIDER_TIMEOUT = 84600
- 特別注意:如果爬蟲在規定時限沒有把request全部爬取完畢,此時強行停止的話,爬取佇列中就還會存有部分request請求。那麼爬蟲下次開始爬取時,一定要記得在master端對爬取佇列進行清空操作。
3.2. 修改scrapy-redis原始碼
# ----------- 修改scrapy-redis原始碼時,特別需要注意的是:---------
# 第一,要留有原始程式碼的備份。
# 第二,當專案移植到其他機器上時,需要將scrapy-redis原始碼一起移植過去。一般程式碼位置在\Lib\site-packages\scrapy_redis\下
- 想象一下,爬蟲已經結束的特徵是什麼?那就是爬取佇列已空,從爬取佇列中無法取到request資訊。那著手點應該就在從爬取佇列中獲取request和排程這個部分。檢視scrapy-redis原始碼,我們發現了兩個著手點:
3.2.1. 細節
# .\Lib\site-packages\scrapy_redis\schedluer.py
def next_request(self):
block_pop_timeout = self.idle_before_close
# 下面是從爬取佇列中彈出request
# 這個block_pop_timeout 我尚未研究清除其作用。不過肯定不是超時時間......
request = self.queue.pop(block_pop_timeout)
if request and self.stats:
self.stats.inc_value('scheduler/dequeued/redis', spider=self.spider)
return request
- 9
# .\Lib\site-packages\scrapy_redis\spiders.py
def next_requests(self):
"""Returns a request to be scheduled or none."""
use_set = self.settings.getbool('REDIS_START_URLS_AS_SET', defaults.START_URLS_AS_SET)
fetch_one = self.server.spop if use_set else self.server.lpop
# XXX: Do we need to use a timeout here?
found = 0
# TODO: Use redis pipeline execution.
while found < self.redis_batch_size:
data = fetch_one(self.redis_key)
if not data:
# 代表爬取佇列為空。但是可能是永久為空,也可能是暫時為空
# Queue empty.
break
req = self.make_request_from_data(data)
if req:
yield req
found += 1
else:
self.logger.debug("Request not made from data: %r", data)
if found:
self.logger.debug("Read %s requests from '%s'", found, self.redis_key)
- 參考註釋,從上述原始碼來看,就只有這兩處可以做手腳。但是爬蟲在爬取過程中,佇列隨時都可能出現暫時的空窗期。想判斷爬取佇列為空,一般是設定一個時限,如果在一個時段內,佇列一直持續為空,那我們可以基本認定這個爬蟲已經結束了。所以有了如下的改動:
# .\Lib\site-packages\scrapy_redis\schedluer.py
# 原始程式碼
def next_request(self):
block_pop_timeout = self.idle_before_close
request = self.queue.pop(block_pop_timeout)
if request and self.stats:
self.stats.inc_value('scheduler/dequeued/redis', spider=self.spider)
return request
# 修改後的程式碼
def __init__(self, server,
persist=False,
flush_on_start=False,
queue_key=defaults.SCHEDULER_QUEUE_KEY,
queue_cls=defaults.SCHEDULER_QUEUE_CLASS,
dupefilter_key=defaults.SCHEDULER_DUPEFILTER_KEY,
dupefilter_cls=defaults.SCHEDULER_DUPEFILTER_CLASS,
idle_before_close=0,
serializer=None):
# ......
# 增加一個計數項
self.lostGetRequest = 0
def next_request(self):
block_pop_timeout = self.idle_before_close
request = self.queue.pop(block_pop_timeout)
if request and self.stats:
# 如果拿到了就恢復這個值
self.lostGetRequest = 0
self.stats.inc_value('scheduler/dequeued/redis', spider=self.spider)
if request is None:
self.lostGetRequest += 1
print(f"request is None, lostGetRequest = {self.lostGetRequest}, time = {datetime.datetime.now()}")
# 100個大概8分鐘的樣子
if self.lostGetRequest > 200:
print(f"request is None, close spider.")
# 結束爬蟲
self.spider.crawler.engine.close_spider(self.spider, 'queue is empty')
return request
- 相關log資訊
2017-12-14 16:18:06 [scrapy.middleware] INFO: Enabled item pipelines:
['redisClawerSlaver.pipelines.beforeRedisPipeline',
'redisClawerSlaver.pipelines.amazonRedisPipeline',
'scrapy_redis.pipelines.RedisPipeline']
2017-12-14 16:18:06 [scrapy.core.engine] INFO: Spider opened
2017-12-14 16:18:06 [scrapy.extensions.logstats] INFO: Crawled 0 pages (at 0 pages/min), scraped 0 items (at 0 items/min)
request is None, lostGetRequest = 1, time = 2017-12-14 16:18:06.370400
request is None, lostGetRequest = 2, time = 2017-12-14 16:18:11.363400
request is None, lostGetRequest = 3, time = 2017-12-14 16:18:16.363400
request is None, lostGetRequest = 4, time = 2017-12-14 16:18:21.362400
request is None, lostGetRequest = 5, time = 2017-12-14 16:18:26.363400
request is None, lostGetRequest = 6, time = 2017-12-14 16:18:31.362400
request is None, lostGetRequest = 7, time = 2017-12-14 16:18:36.363400
request is None, lostGetRequest = 8, time = 2017-12-14 16:18:41.362400
request is None, lostGetRequest = 9, time = 2017-12-14 16:18:46.363400
request is None, lostGetRequest = 10, time = 2017-12-14 16:18:51.362400
2017-12-14 16:18:56 [scrapy.core.engine] INFO: Closing spider (queue is empty)
request is None, lostGetRequest = 11, time = 2017-12-14 16:18:56.363400
request is None, close spider.
登入結果:loginRes = (235, b'Authentication successful')
登入成功,code = 235
mail has been send successfully. message:Content-Type: text/plain; charset="utf-8"
MIME-Version: 1.0
Content-Transfer-Encoding: base64
From: [email protected]
To: [email protected]
Subject: =?utf-8?b?54is6Jmr57uT5p2f54q25oCB5rGH5oql77yabmFtZSA9IHJlZGlzQ2xhd2VyU2xhdmVyLCByZWFzb24gPSBxdWV1ZSBpcyBlbXB0eSwgZmluaXNoZWRUaW1lID0gMjAxNy0xMi0xNCAxNjoxODo1Ni4zNjQ0MDA=?=
57uG6IqC77yacmVhc29uID0gcXVldWUgaXMgZW1wdHksIHN1Y2Nlc3NzISBhdDoyMDE3LTEyLTE0
IDE2OjE4OjU2LjM2NDQwMA==
2017-12-14 16:18:56 [scrapy.statscollectors] INFO: Dumping Scrapy stats:
{'finish_reason': 'queue is empty',
'finish_time': datetime.datetime(2017, 12, 14, 8, 18, 56, 364400),
'log_count/INFO': 8,
'start_time': datetime.datetime(2017, 12, 14, 8, 18, 6, 362400)}
2017-12-14 16:18:56 [scrapy.core.engine] INFO: Spider closed (queue is empty)
Unhandled Error
Traceback (most recent call last):
File "E:\Miniconda\lib\site-packages\scrapy\commands\runspider.py", line 89, in run
self.crawler_process.start()
File "E:\Miniconda\lib\site-packages\scrapy\crawler.py", line 285, in start
reactor.run(installSignalHandlers=False) # blocking call
File "E:\Miniconda\lib\site-packages\twisted\internet\base.py", line 1243, in run
self.mainLoop()
File "E:\Miniconda\lib\site-packages\twisted\internet\base.py", line 1252, in mainLoop
self.runUntilCurrent()
--- <exception caught here> ---
File "E:\Miniconda\lib\site-packages\twisted\internet\base.py", line 878, in runUntilCurrent
call.func(*call.args, **call.kw)
File "E:\Miniconda\lib\site-packages\scrapy\utils\reactor.py", line 41, in __call__
return self._func(*self._a, **self._kw)
File "E:\Miniconda\lib\site-packages\scrapy\core\engine.py", line 137, in _next_request
if self.spider_is_idle(spider) and slot.close_if_idle:
File "E:\Miniconda\lib\site-packages\scrapy\core\engine.py", line 189, in spider_is_idle
if self.slot.start_requests is not None:
builtins.AttributeError: 'NoneType' object has no attribute 'start_requests'
2017-12-14 16:18:56 [twisted] CRITICAL: Unhandled Error
Traceback (most recent call last):
File "E:\Miniconda\lib\site-packages\scrapy\commands\runspider.py", line 89, in run
self.crawler_process.start()
File "E:\Miniconda\lib\site-packages\scrapy\crawler.py", line 285, in start
reactor.run(installSignalHandlers=False) # blocking call
File "E:\Miniconda\lib\site-packages\twisted\internet\base.py", line 1243, in run
self.mainLoop()
File "E:\Miniconda\lib\site-packages\twisted\internet\base.py", line 1252, in mainLoop
self.runUntilCurrent()
--- <exception caught here> ---
File "E:\Miniconda\lib\site-packages\twisted\internet\base.py", line 878, in runUntilCurrent
call.func(*call.args, **call.kw)
File "E:\Miniconda\lib\site-packages\scrapy\utils\reactor.py", line 41, in __call__
return self._func(*self._a, **self._kw)
File "E:\Miniconda\lib\site-packages\scrapy\core\engine.py", line 137, in _next_request
if self.spider_is_idle(spider) and slot.close_if_idle:
File "E:\Miniconda\lib\site-packages\scrapy\core\engine.py", line 189, in spider_is_idle
if self.slot.start_requests is not None:
builtins.AttributeError: 'NoneType' object has no attribute 'start_requests'
Process finished with exit code 0
- 有一個問題,如上所述,當通過engine.close_spider(spider, ‘reason’)來關閉spider時,有時會出現幾個錯誤之後才能關閉。可能是因為scrapy會開啟多個執行緒同時抓取,然後其中一個執行緒關閉了spider,其他執行緒就找不到spider才會報錯。
3.2.2. 注意事項
整個排程過程如下:
-
scheduler.py
-
queue.py
- 所以,PriorityQueue和另外兩種佇列FifoQueue,LifoQueue有所不同,特別需要注意。如果會使用到timeout這個引數,那麼在setting中就只能指定爬取佇列為FifoQueue或LifoQueue。
# 指定排序爬取地址時使用的佇列,
# 預設的 按優先順序排序(Scrapy預設),由sorted set實現的一種非FIFO、LIFO方式。
# 'SCHEDULER_QUEUE_CLASS': 'scrapy_redis.queue.SpiderPriorityQueue',
# 可選的 按先進先出排序(FIFO)
'SCHEDULER_QUEUE_CLASS': 'scrapy_redis.queue.SpiderQueue',
# 可選的 按後進先出排序(LIFO)
# SCHEDULER_QUEUE_CLASS = 'scrapy_redis.queue.SpiderStack'