Jochen, thanks for you reply!

As I mentioned in my first post, I can see with tcpdump that packet lenght, 
that comes to Graylog node is more than 8192 bytes:
$ sudo tcpdump -n -i tunl0 port 12500 and udp
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on tunl0, link-type RAW (Raw IP), capture size 65535 bytes
18:22:19.062304 IP 10.218.52.42.56745 > 10.218.50.20.12500: UDP, length 54
18:22:19.079891 IP 10.218.52.42.56745 > 10.218.50.20.12500: UDP, length 81
18:22:19.113119 IP 10.218.52.42.56745 > 10.218.50.20.12500: UDP, length 92
18:22:19.117398 IP 10.218.52.42.56745 > 10.218.50.20.12500: UDP, length 93
18:22:19.121636 IP 10.218.52.42.56745 > 10.218.50.20.12500: UDP, length 81
18:22:19.123707 IP 10.218.52.42.56745 > 10.218.50.20.12500: UDP, length 94
18:22:22.092734 IP 10.218.49.5.61843 > 10.218.50.20.12500: UDP, length 108
18:22:22.093300 IP 10.218.49.5.61843 > 10.218.50.20.12500: UDP, length 70
18:22:22.238882 IP 10.218.49.5.61843 > 10.218.50.20.12500: UDP, length 71
18:22:24.067068 IP 10.218.49.4.59298 > 10.218.50.20.12500: UDP, length 87
18:22:26.148394 IP 10.218.52.42.56745 > 10.218.50.20.12500: UDP, length 85
18:22:27.477703 IP 10.218.49.4.59298 > 10.218.50.20.12500: UDP, length 13642
18:22:31.158020 IP 10.218.52.42.56745 > 10.218.50.20.12500: UDP, length 91
18:22:35.945376 IP 10.218.49.6.63104 > 10.218.50.20.12500: UDP, length 69
18:22:35.945489 IP 10.218.49.6.63104 > 10.218.50.20.12500: UDP, length 91
18:22:37.279499 IP 10.218.49.5.61843 > 10.218.50.20.12500: UDP, length 108

Also, I have another node with Graylog2 0.20, messages from same sources 
comes to logstash UDP input without problems.

So, I don't thinks that it is network related problems.

On Thursday, May 7, 2015 at 4:04:26 PM UTC+3, Jochen Schalanda wrote:
>
> Hi Arkadiy,
>
> not all network devices support UDP packets bigger than 8KiB (8192 bytes) 
> and this seems to be the case on your network somewhere. To circumvent this 
> restriction, Graylog or more specifically the GELF format supports chunking 
> which means splitting a large message into multiple UDP packets of a 
> certain size (see https://www.graylog.org/resources/gelf for details on 
> the GELF format). So in order to solve your problem you could collect the 
> logs on the same machine which produces those raw log lines with a log 
> shipper like nxlog or logstash and send them to Graylog via GELF. 
> Alternatively you'll have to switch from UDP to TCP which doesn't suffer 
> this kind of restriction.
>
> Cheers,
> Jochen
>
> On Thursday, 7 May 2015 13:56:25 UTC+2, Arkadiy Shinkarev wrote:
>>
>> Hi!
>>
>> I'm trying to send messages to raw udp input in Graylog 1.0.2.
>> The message size is 1k-20k, but Graylog only shown first 8k of message.
>>
>> I have configured "recv_buffer_size: 10485760" for input, also 
>> set net.core.rmem_max = 26214400 in sysctl.conf.
>>
>> When I run tcpdump, I see that message len is ok (>8k).
>> When I run strace -e trace=network i see the following:
>> [pid 10539] recvfrom(365, 
>> "\24\21\0\0\370\370\352\1778q\1\0\1\0\0\0\10\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"...,
>>  
>> 8192, 0, NULL, NULL) = 4372
>> [pid 10539] recvfrom(365, 
>> "\24\21\0\0\371\370\352\1779q\1\0\1\0\0\0\10\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"...,
>>  
>> 8192, 0, NULL, NULL) = 4372
>> [pid 10537] <... recvfrom resumed> 
>> "\24\21\0\0O\372\352\177\235q\1\0\1\0\0\0\10\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"...,
>>  
>> 8192, 0, NULL, NULL) = 4372
>> [pid 10537] recvfrom(365, 
>> "\24\21\0\0X\372\352\177\236q\1\0\1\0\0\0\10\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"...,
>>  
>> 8192, 0, NULL, NULL) = 4372
>> [pid 10541] recvfrom(365, 
>> "\24\21\0\0\200\372\352\177\240q\1\0\1\0\0\0\10\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"...,
>>  
>> 8192, 0, NULL, NULL) = 4372
>> [pid 10541] recvfrom(365, 
>> "\24\21\0\0\201\372\352\177\241q\1\0\1\0\0\0\10\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"...,
>>  
>> 8192, 0, NULL, NULL) = 4372
>> [pid 10534] recvfrom(365, 
>> "\24\21\0\0\302\372\352\177\330q\1\0\1\0\0\0\10\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"...,
>>  
>> 8192, 0, NULL, NULL) = 4372
>> [pid 10534] recvfrom(365, 
>> "\24\21\0\0\303\372\352\177\331q\1\0\1\0\0\0\10\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"...,
>>  
>> 8192, 0, NULL, NULL) = 4372
>> [pid 10540] <... recvfrom resumed> 
>> "\24\21\0\0$\373\352\177-r\1\0\1\0\0\0\10\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 
>> 8192, 0, NULL, NULL) = 4372
>> [pid 10540] <... recvfrom resumed> 
>> "\24\21\0\0&\373\352\177/r\1\0\1\0\0\0\10\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 
>> 8192, 0, NULL, NULL) = 4372
>> [pid 10535] recvfrom(362, 
>> "\24\21\0\0b\373\352\177jr\1\0\1\0\0\0\10\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 
>> 8192, 0, NULL, NULL) = 4372
>> [pid 10535] recvfrom(362, 
>> "\24\21\0\0c\373\352\177kr\1\0\1\0\0\0\10\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 
>> 8192, 0, NULL, NULL) = 4372
>> [pid 10537] recvfrom(362, 
>> "\24\21\0\0\315\375\352\177_s\1\0\1\0\0\0\10\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"...,
>>  
>> 8192, 0, NULL, NULL) = 4372
>> [pid 10537] recvfrom(362, 
>> "\24\21\0\0\316\375\352\177`s\1\0\1\0\0\0\10\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"...,
>>  
>> 8192, 0, NULL, NULL) = 4372
>> [pid 10538] recvfrom(362, 
>> "\24\21\0\0\v\376\352\177\224s\1\0\1\0\0\0\10\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"...,
>>  
>> 8192, 0, NULL, NULL) = 4372
>> [pid 10538] recvfrom(362, 
>> "\24\21\0\0\f\376\352\177\225s\1\0\1\0\0\0\10\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"...,
>>  
>> 8192, 0, NULL, NULL) = 4372
>> [pid 10534] <... recvfrom resumed> 
>> "\24\21\0\0\234\376\352\177\30t\1\0\1\0\0\0\10\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"...,
>>  
>> 8192, 0, NULL, NULL) = 4372
>> [pid 10534] <... recvfrom resumed> 
>> "\24\21\0\0\237\376\352\177\33t\1\0\1\0\0\0\10\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"...,
>>  
>> 8192, 0, NULL, NULL) = 4372
>>
>> I'm also test GELF UDP input, messages comes from logstash (to logstash 
>> messages comes from UDP input with 32k buffer size), there is no problem - 
>> message looks good, Graylog shows full message.
>>
>> Where is the problem?
>>
>> Some additional information:
>> OS: CentOS release 6.5 (Final)
>> Kernel: 2.6.32-431.29.2.el6.centos.plus.x86_64
>> Graylog: 1.0.2
>>
>> 2 graylog-server nodes behind load balancer (LVS) + 2 nodes elasticsearch 
>> cluster.
>>
>> Thank you!
>>
>

-- 
You received this message because you are subscribed to the Google Groups 
"graylog2" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to