Also, I checked that I have http.js in the following modules and none of
them have setTimeout (did a grep).
./express/lib/http.js
./express/node_modules/connect/lib/http.js
./socket.io/lib/transports/http.js
On Tuesday, 6 March 2012 22:56:42 UTC+5:30, Arindam wrote:
>
> I am using node 0.6.10 version and I got the error today while uploading a
> file to amazon S3.
> I am using knox(@0.0.9) module with request(@2.9.153).
>
> [2012-03-06 22:33:34.251] [ERROR] choicebeat - Error uploading to S3:
> Error: soc
> ket hang up
> at createHangUpError (http.js:1104:15)
> at Socket.<anonymous> (http.js:1207:27)
> at Socket.emit (events.js:88:20)
> at Array.0 (net.js:320:10)
> at EventEmitter._tickCallback (node.js:192:40)
>
> Looks like the issue was fixed. Am I using any older version ?
>
> On Thursday, 23 June 2011 05:16:06 UTC+5:30, murvinlai wrote:
>>
>> You can try (temporary ) edit the node.js library. lib\http.js
>> search for socket.setTimeout the default timeout should be 2 * 60 *
>> 1000 = 2 minutes , instead of 20 seconds. You can change that value.
>>
>> however, if your remote client has a slow response, then you will get a
>> lot of timeout.
>>
>> Also, a warning is that, if you let your process run for long time, and
>> the socket opened (for connection pool) is over your maxSockets, you may
>> not be able to establish any new http.request.
>>
>> before a formal fix, I suggest the followings:
>> 1) increase your maxSocket (and also your OS one. ulimit -a and use
>> ulimit -n to increase it , or change the /etc/security/limits.conf
>> Incraese soft and hard. (where hard > soft). e.g * soft nofile 10000
>> * hard nofile 12000
>>
>> 2) catch http.request error. i.e.:
>> var clientRequest = http.request (... );
>> clientRequest.on ('error', function() { clientRequest.abort() } );
>>
>> Although socket timeout will kill and free up the socket, but I find out
>> doing that will save you more socket.
>>
>> 3) lower your socket timeout. change the http.js and then "Make" again.
>>
>>
>> On Wed, Jun 22, 2011 at 4:06 PM, Kiran <[email protected]> wrote:
>>
>>> Has anybody else solved this problem. I am having a similar issue.
>>> When I am using Apache Bench to test my node server connecting to a
>>> remote server, i occasionally get a hang of the entire Node process
>>> for 20 seconds. THe number 20 is key because it is the default socket
>>> timeout in Node so somehow having a bad socket is causing a real
>>> problem with nodes ability to be fully async and performant. Is there
>>> a way to reduce the timeout and/or is there a way to get this issue to
>>> be resolved or at least a patch to the native code that shows were
>>> this might be a problem.
>>>
>>> Kiran
>>>
>>> On Jun 20, 2:27 pm, murvinlai <[email protected]> wrote:
>>> > I haven't heard any news yet. Here is my post in github:
>>> https://gist.github.com/1032413
>>> > however, I get one of the core developers to take a look into this
>>> > problem. :)
>>> >
>>> > What is "no route to host"?
>>> >
>>> > Please see my code in the link I provide. I have a very controlled
>>> > testing. I think when doing remote call, no matter how many socket
>>> > you set (e.g. maxSockets= 1000), socket will break /hangupbecause
>>> > of the client's side response being timeout by whatever reason. Then
>>> > once that socket has problem, it will never be recovered. If you
>>> useupall the good sockets, then your program cannot make any
>>> > http.request and you can declare your node die.
>>> >
>>> > There is a get around solution I'm using before they comeupwith the
>>> > correct solution.
>>> > 1) use Agent, and set Agent.maxSockets = big number that matches with
>>> > your ulimit. And http.options.headers.Connection:'keep-alive'
>>> > 2) Make your own timeout route. If it times out, make sure to call
>>> > http.clientRequest.abort(); But, your socket still die, just that it
>>> > won't timeout and the socket timeout again.
>>> > 3) When timeout or on.('error') with error message is "sockethangup"
>>> > more than certain number of times, (e.g. 100), then kill and restart
>>> > node. I don't know if just restart http.server will work, but I'm
>>> > sure restarting node will work.
>>> > 4) To restart node inside code, I use pgte/fugue .
>>> https://github.com/pgte/fugue
>>> > I have done some testings and it looks good.
>>> >
>>> > This workaround seems to be promising. But, if you use it, please
>>> > let me know if you find out anything. :) Thanks.
>>> >
>>> > On Jun 18, 11:55 pm, Luca922 <arne%[email protected]>
>>> > wrote:
>>> >
>>> >
>>> >
>>> >
>>> >
>>> >
>>> >
>>> > > Any news on this issue?
>>> >
>>> > > The sockethangupseems to happen on 0.5pre as well, usually in a
>>> > > situation where a "no route to host" occurs.
>>> >
>>> > > The problem can be reproduced by us as we have one EC2 box constantly
>>> > > doing http.requests to our own server. This worked perfectly fine
>>> > > until we replaced our server with a more powerful box in a different
>>> > > data center. For some reason, our monitoring reports that - only on
>>> > > two or three requests a day - the EC2 cannot find the route to our
>>> new
>>> > > server. No problems like that with the old server. Anyway, this is
>>> > > when Node gets stuck.
>>> >
>>> > > Luca
>>>
>>> --
>>> You received this message because you are subscribed to the Google
>>> Groups "nodejs" group.
>>> To post to this group, send email to [email protected].
>>> To unsubscribe from this group, send email to
>>> [email protected].
>>> For more options, visit this group at
>>> http://groups.google.com/group/nodejs?hl=en.
>>>
>>>
>>
--
Job Board: http://jobs.nodejs.org/
Posting guidelines:
https://github.com/joyent/node/wiki/Mailing-List-Posting-Guidelines
You received this message because you are subscribed to the Google
Groups "nodejs" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to
[email protected]
For more options, visit this group at
http://groups.google.com/group/nodejs?hl=en?hl=en
