I wrote a epoll wrapper in python, It works fine but recently I found the performance is not not ideal for large package sending. I look down into the cod开发者_StackOverflowe and found there's actually a LOT of error
Traceback (most recent call last):
File "/Users/dawn/Documents/workspace/work/dev/server/sandbox/single_point/tcp_epoll.py", line 231, in send_now
num_bytes = self.sock.send(self.response)
error: [Errno 35] Resource temporarily unavailable
and previously silent it as the document said, so my sending function was done this way:
def send_now(self):
'''send message at once'''
st = time.time()
times = 0
while self.response != '':
try:
num_bytes = self.sock.send(self.response)
l.info('msg wrote %s %d : %r size %r',self.ip,self.port,self.response[:num_bytes],num_bytes)
self.response = self.response[num_bytes:]
except socket.error,e:
if e[0] in (errno.EWOULDBLOCK,errno.EAGAIN):
#here I printed it, but I silent it in normal days
#print 'would block, again %r',tb.format_exc()
break
else:
l.warning('%r %r socket error %r',self.ip,self.port,tb.format_exc())
#must break or cause dead loop
break
except:
#other exceptions
l.warning('%r %r msg write error %r',self.ip,self.port,tb.format_exc())
break
times += 1
et = time.time()
I googled it, and says it caused by temporarily network buffer run out So how can I manually and efficiently detect this error instead it goes to exception phase?
Because it cause to much time to rasie/handle the exception.
Use select
to see if a socket is ready for reading or writing.
精彩评论