[sr-dev] Big TCP message
Amit Elbaz
amit at blueseasoft.co.il
Sun Jan 6 16:08:09 CET 2013
Thanks. your suggestions worked!
Amit.
On Thu, Jan 3, 2013 at 5:15 PM, Klaus Darilion <klaus.mailinglists at pernau.at
> wrote:
> Ups.
>
>
> On 03.01.2013 15:11, Klaus Darilion wrote:
>
>> How much traffic do you want to send?
>>
>> When does the problem happen? At which message size does it happen?
>>
>> I just wonder that the normal options which works for us do not work for
>> you and t
>>
> and the problem is code related.
>
> 1. Check if Kamailio really uses the buffer size you have configured (via
> sercmd config framework):
> # kamctl sercmd cfg.get tcp rd_buf_size
>
> 2. Test on which message size the problems start. Attached is a perl
> script which allows you to generate OPTIONS messages with a specified
> message size, e.g:
>
> # ./send-tcp-message.pl 127.0.0.1:5060 2048
>
>
> regards
> Klaus
>
>
>> regards
>> Klaus
>>
>> On 03.01.2013 14:21, Amit Elbaz wrote:
>>
>>> I increased tcp_rd_buf_size, but it did not help.
>>> I don't think this is the problem.
>>>
>>> In tcp_read.c:
>>> int tcp_read(struct tcp_connection *c, int* flags)
>>> {
>>> int bytes_free, bytes_read;
>>> struct tcp_req *r;
>>> int fd;
>>>
>>> r=&c->req;
>>> fd=c->fd;
>>> bytes_free=r->b_size- (int)(r->pos - r->buf);
>>> if (unlikely(bytes_free==0)){
>>> LOG(L_ERR, "ERROR: tcp_read: buffer overrun, dropping\n");
>>> r->error=TCP_REQ_OVERRUN;
>>> return -1;
>>> }
>>>
>>> how to i increase r->b_size ?
>>>
>>> Thanks for you help, Amit.
>>>
>>> On Wed, Jan 2, 2013 at 6:51 PM, Klaus Darilion
>>> <klaus.mailinglists at pernau.at <mailto:klaus.mailinglists@**pernau.at<klaus.mailinglists at pernau.at>
>>> >>
>>> wrote:
>>>
>>> Hi!
>>>
>>> -b is the receive buffer of the socket in the OS (for UDP tweaking,
>>> I do not know if this is relevant for TCP), so not related to this
>>> problem as this problem happens inside Kamailio.
>>>
>>> Actually 15K is rather small, really strange that you get errors.
>>> So, I did some tests:
>>>
>>> 1. tcp_rd_buf_size must be increased. The documentation is bad as it
>>> meantions "TCP datagram" but actually means the maximum size of SIP
>>> (or HTTP) message that can be received.
>>>
>>> 2. tcp_rd_buf_size is internally limited to 16MBytes. If you need
>>> more, you have to change the code in:
>>> tcp_options.c: { "rd_buf_size", CFG_VAR_INT | CFG_ATOMIC, 512,
>>> 16777216, 0, 0,
>>>
>>> 3. If you use real large message you may need to increase the
>>> private memory, and if you forward it with tm module, also the
>>> shared memory.
>>>
>>> regards
>>> Klaus
>>>
>>>
>>> Am 02.01.2013 13:59, schrieb Amit Elbaz:
>>>
>>>> this still does not solve the problem.
>>>> I've also tried running Kamailio with *-b 32768* option.
>>>>
>>>> here are the logs:
>>>>
>>>> Jan 2 14:54:24 debdev kamailio[2580 <tel:%5B2580>]: ERROR: <core>
>>>> [tcp_read.c:344]: ERROR: tcp_read: buffer overrun, dropping
>>>> Jan 2 14:54:24 debdev kamailio[2580 <tel:%5B2580>]: ERROR: <core>
>>>> [tcp_read.c:1127]: ERROR: tcp_read_req: error reading
>>>>
>>>> Thanks for your help.
>>>>
>>>> On Wed, Jan 2, 2013 <tel:2013> at 1:17 PM, Andrew Pogrebennyk
>>>> <apogrebennyk at sipwise.com <mailto:apogrebennyk at sipwise.**com<apogrebennyk at sipwise.com>>>
>>>> wrote:
>>>>
>>>> Hi,
>>>> you may try changing the tcp_rd_buf_size and tcp_conn_wq_max:
>>>>
>>>> http://www.kamailio.org/wiki/**cookbooks/3.3.x/core#tcp_conn_**wq_max<http://www.kamailio.org/wiki/cookbooks/3.3.x/core#tcp_conn_wq_max>
>>>>
>>>> And since this in development mailing list I'd like to point
>>>> at what
>>>> seem an iaccuracy in documentation: several options refer to
>>>> the core
>>>> parameter called tcp_write_buf, which isn't described in the
>>>> cookbook:
>>>>
>>>> tcp_conn_wq_max
>>>> Maximum bytes queued for write allowed per connection.
>>>> Attempting to
>>>> queue more bytes would result in an error and in the
>>>> connection being
>>>> closed (too slow). If *tcp_write_buf* is not enabled, it has
>>>> no effect.
>>>>
>>>> I hope Daniel or someone who knows reads this..
>>>>
>>>> BR,
>>>> Andrew
>>>>
>>>> On 01/02/2013 10:38 AM, Amit Elbaz wrote:
>>>> > Hi,
>>>> >
>>>> > While sending a big TCP SIP message (15K+), I get "tcp_read:
>>>> buffer
>>>> > overrun, dropping" message.
>>>> > this happens after a few fragments have been sent, and then
>>>> i get this
>>>> > message and the connection resets.
>>>> > is it possible to set the max-buffer size through
>>>> configuration? if so,
>>>> > where?
>>>> >
>>>> > Thanks, Amit.
>>>> >
>>>> >
>>>> > ______________________________**_________________
>>>> > sr-dev mailing list
>>>> > sr-dev at lists.sip-router.org
>>>> <mailto:sr-dev at lists.sip-**router.org <sr-dev at lists.sip-router.org>>
>>>> > http://lists.sip-router.org/**cgi-bin/mailman/listinfo/sr-**
>>>> dev <http://lists.sip-router.org/cgi-bin/mailman/listinfo/sr-dev>
>>>> >
>>>>
>>>>
>>>> ______________________________**_________________
>>>> sr-dev mailing list
>>>> sr-dev at lists.sip-router.org <mailto:sr-dev at lists.sip-**
>>>> router.org <sr-dev at lists.sip-router.org>>
>>>> http://lists.sip-router.org/**cgi-bin/mailman/listinfo/sr-**dev<http://lists.sip-router.org/cgi-bin/mailman/listinfo/sr-dev>
>>>>
>>>>
>>>>
>>>>
>>>> ______________________________**_________________
>>>> sr-dev mailing list
>>>> sr-dev at lists.sip-router.org <mailto:sr-dev at lists.sip-**router.org<sr-dev at lists.sip-router.org>
>>>> >
>>>> http://lists.sip-router.org/**cgi-bin/mailman/listinfo/sr-**dev<http://lists.sip-router.org/cgi-bin/mailman/listinfo/sr-dev>
>>>>
>>>
>>>
>>>
>> ______________________________**_________________
>> sr-dev mailing list
>> sr-dev at lists.sip-router.org
>> http://lists.sip-router.org/**cgi-bin/mailman/listinfo/sr-**dev<http://lists.sip-router.org/cgi-bin/mailman/listinfo/sr-dev>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.sip-router.org/pipermail/sr-dev/attachments/20130106/9b41c3b9/attachment-0001.htm>
More information about the sr-dev
mailing list