I Have this strange networking problem.
Afer 36 hours (sometimes more sometimes less) of system runtime, loading web pages on any browser becomes painfully slow or completely impossibe.
When this happens I can still ping to IP addresses or their regular addresses.
I tried to restart networking service, take eth0 interface down and back up, I even moved from 32bit to 64bit on a completely different hardware.
So far my only solution is rebooting.
# uname -a
Linux Megatron 3.4.2-2-ARCH #1 SMP PREEMPT Mon Jun 11 22:27:17 CEST 2012 x86_64 GNU/Linux
Last edited by GlockX (2012-10-24 09:33:06)
I suggest you log some traffic when browsing becomes slow again. That way, you might be able to see what's the cause:
- Lots of other traffic
- Slow DNS traffic
- Lots of renegotiations (because of a bad line / crowded radio frequency)
At first I thought something might be poisoning your routing tables, but that would have the same effect on ICMP.
Thank you for your reply!
Unfortunately I havent been able to resolve this. I checked the log files and didn't see anything suspicious, I also tried to change DNS servers.
lately I need to reboot my pc every 24 hours because of this.
I don't want to send you in the wrong direction but I had a similar problem and contacted my isp. They began logging my connection and after seeing there was a problem asked me to switch off my router at night. Their explanation was that the quality of internet connections can degrade the longer they are maintained. It did initially resolve the problem but I've found that as time goes on I don't need to switch the router off as often.
Hi Larry, thank you for your reply.
I have other computers (Windows and Mac) at home and they work just fine, so I don't think its a router nor an isp problem.
Before Arch I had Debian on that 32bit computer for 6 years without this annoying issue, I think something is wrong with my system configuration but I can't figure out what exactly.
I even moved from 32bit to 64bit on a completely different hardware.
What exactly did stay the same during your experiments? Assuming that you mean you had the same problem on different bitnesses of Arch, on different machines, but with the same router, I can only think of either a faulty cable between the router and your workspace, or some problem with the router's (hardware) port you are connecting to.
Are you sure that when you capture your traffic while having problems, the same patterns show as when you're not having problems, only with greater delays?
After some research with Wireshark I noticed this error message: "TCP Previous segment not captured", which led me to this page:
http://ask.wireshark.org/questions/1114 … -rare-case
It seems that Arch's default MTU size is 576 so I changed it to 1492, and it kind of fixed the problem. everything works except the most important thing every internet user must have access to: Google.
I tried all kind of MTU values in the 1400's ,1500 exactly, or even higher values but my browser is having trouble loading up Google pages.
If I set it too low then other pages won't load.
BTW, My router's MTU is 1492
It seems that Arch's default MTU size is 576 so I changed it to 1492, and it kind of fixed the problem.
First to clarify, there is no such thing as a default MTU in Arch or <insert your distro here>. Likely dhcp is setting the wrong MTU size or the network driver you are using in the kernel is having issues with your hardware.
How are you obtaining an IPv4 address (dhcp or static)?
The default MTU for Ethernet is 1500 bytes.
To troubleshoot, set your MTU to 1500
ifconfig eth0 mtu 1500
Ping your gateway (change gw ip address accordingly)
ping -M do -s 1472 192.168.1.1
If you get
icmp_seq=1 Frag needed and DF set (mtu = 1500)
Decrease 1472 until you find the highest value that succeeds
tux:~ $ ping -M do -s 1472 10.20.2.1
PING 10.20.2.1 (10.20.2.1) 1472(1500) bytes of data.
1480 bytes from 10.20.2.1: icmp_req=1 ttl=64 time=0.497 ms
Report back your results. This should give us a better idea to figure out what is going on.
Last edited by pyther (2012-09-03 21:01:13)
Here are the results:
[root@Megatron ~]# ping -M do -s 1472 192.168.1.1 PING 192.168.1.1 (192.168.1.1) 1472(1500) bytes of data. 1480 bytes from 192.168.1.1: icmp_req=1 ttl=64 time=1.25 ms 1480 bytes from 192.168.1.1: icmp_req=2 ttl=64 time=1.16 ms ^C --- 192.168.1.1 ping statistics --- 2 packets transmitted, 2 received, 0% packet loss, time 1001ms rtt min/avg/max/mdev = 1.167/1.210/1.253/0.043 ms
if I ping my gateway with 1473 bytes, It fails:
From 192.168.1.100 icmp_seq=1 Frag needed and DF set (mtu = 1500)
[root@Megatron ~]# lspci | grep net 01:00.0 Ethernet controller: Realtek Semiconductor Co., Ltd. RTL8111/8168B PCI Express Gigabit Ethernet controller (rev 06)
using the r8169 module. if I remember right the same module used to be loaded on my old PC(Gigabit MB also) which had the same issue.
I blacklisted and replaced r8169 with r8168, hope it'll give some better results. I'll know the answer in about 30 hours.
Last edited by GlockX (2012-09-04 00:31:20)
Unfortunately, with MTU set to 1500 and the r8168 module Google still doesn't load properly . sometimes it doesn't load at all.
This is interesting, i had always problems with my internet connection on my pc and now i changed the mtu to 1480 and my internet connection had a speedup. Websites are loading faster, if i change it back i can measure the difference. Really cool. How does this work? How can this be? I thought this doesn't make any difference?!
I am an Ubuntu user (don't ban me). I have the exactly same problem. Tried everything! It doesn't work neither in Ubuntu neither in Fedora. My card is Realtek Semiconductor Co., Ltd. RTL8101E/RTL8102E PCI Express Fast Ethernet controller (rev 02) and uses that damned r8169 driver. I tried using a newer driver from the realtek website (the 8101 version) but I stil have the same problem.
I cannot connect to anything related to google or to google itself. For example if I go to phoronix.com the site loads but I get a connecting to ajax.googleapis.com that stays that way for a few minutes before the site loads completely.
Any ideas on what to do? It just seems that after 36 hours the network stack gets mangled and starts losing segments. I don't have this problem in windoze. It is not related to DNS or to mtu (tried everything here also). I also tried disabling the firewall, looking at route -n nothing wrong here still doesn't work.
What I did notice is that if I get rid of the router and use pppoeconf to connect directly (I am on a pppoe connection) then it works, but the moment I put the router back and I connect with wired network eth0 I get the behavior back (cannot go near anything google).
Another interesting thing is that if I cannot use google but I can use startpage.com (that uses google). What could it be? Some tracking cookies something that google uses but startpage evades?
Problem solved by using the LTS kernel 3.0.46.