TCP sessions timing out?

Connection sharing, Firewall, Samba..etc
Forum rules
Before you post please read how to get help
Post Reply
leo.smith
Level 1
Level 1
Posts: 24
Joined: Wed Jan 02, 2013 12:23 pm

TCP sessions timing out?

Post by leo.smith » Thu Sep 18, 2014 9:14 am

I have had two networking issues with Mint 17 and the new kernel that didnt happen with mint 14. I cant seem to find them mentioned anywhere at all, except possibly in some vague upstream bug.

One I have been able to work around, the other persists.

BUG #1. I have a virtual private server hosted out in a machine room somewhere and I use ssh to connect to it. Normally I leave a session running their permanently, usually to tail a mail log file to look for mail spam incoming and adjust filter rules. On MINT 14 that would stay up forever, but in 17 it was freezing and needing to be restarted if the data stream was idle for any period of time. Now it might be the NAT router or it might not, but all that has changed in the desktop OS and it was fine the day before I upgraded and not fine the day after.

This has been worked around by using ssh keepalives to continually exchange data between client and server. This has fixed the problem.

In the .ssh dir under my home dir


~/.ssh $ more config
Host vps.xxxx.co.uk
User myname
ServerAliveInterval 60

(data obfuscated for privacy).

Bug #2. I run Thunderbird (31.1.1) for mail and news. News is from the albasani news server and its set to poll every few minutes for 'new news' It does this fine. However that's only the news headers. To download the actual posts, it seems to keep open a second connection. That one dies.

When I decide to read more news, the first time I access the news part of thunderbird to get the actual posts, it results in a 'connection refused' error.

Until I try with a different post, which then succeeds.

As do all subsequent ones until I leave it idle....

--------------------

BOTH of these symptoms are consistent with idle TCP sessions being silently dropped within the kernel. to the point where attempts to use them for data transfer result on some kind of error, which in the case of thunderbird is interpreted as 'unable to connect' and in the case of ssh as no response at all.

Can anyone else confirm similar responses? Or shed any light on this?

It is only an irritation, but it is nevertheless an irritant!

User avatar
PatH57
Level 13
Level 13
Posts: 4576
Joined: Tue Mar 25, 2014 12:11 pm
Location: here and there

Re: TCP sessions timing out?

Post by PatH57 » Thu Sep 18, 2014 9:24 am

Hi,

is this happening on ethernet or wifi?
also share your:

Code: Select all

ifconfig

Code: Select all

sudo lshw -c network
People disagree with me. I just ignore them.
(Linus Torvalds, regarding the use of C++ for the Linux kernel.)

Please Add [Solved] to the topic-title of your first post when appropriate so others know they might find a solution here.

leo.smith
Level 1
Level 1
Posts: 24
Joined: Wed Jan 02, 2013 12:23 pm

Re: TCP sessions timing out?

Post by leo.smith » Thu Sep 18, 2014 10:26 am

Oh all ethernet. Don't do wifi on the desktop.

Code: Select all

$ ifconfig
eth0      Link encap:Ethernet  HWaddr 00:a0:cc:7c:31:27  
          inet addr:192.168.0.6  Bcast:192.168.0.255  Mask:255.255.255.0
          inet6 addr: fe80::2a0:ccff:fe7c:3127/64 Scope:Link
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:90425 errors:0 dropped:0 overruns:0 frame:0
          TX packets:82809 errors:1 dropped:0 overruns:1 carrier:1
          collisions:0 txqueuelen:1000 
          RX bytes:77313566 (77.3 MB)  TX bytes:16517156 (16.5 MB)

lo        Link encap:Local Loopback  
          inet addr:127.0.0.1  Mask:255.0.0.0
          inet6 addr: ::1/128 Scope:Host
          UP LOOPBACK RUNNING  MTU:65536  Metric:1
          RX packets:21238 errors:0 dropped:0 overruns:0 frame:0
          TX packets:21238 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:0 
          RX bytes:11952059 (11.9 MB)  TX bytes:11952059 (11.9 MB)
-------------------
 *-network               
       description: Ethernet interface
       product: DP83815 (MacPhyter) Ethernet Controller
       vendor: National Semiconductor Corporation
       physical id: 0
       bus info: pci@0000:07:00.0
       logical name: eth0
       version: 00
       serial: 00:a0:cc:7c:31:27
       size: 100Mbit/s
       capacity: 100Mbit/s
       width: 32 bits
       clock: 33MHz
       capabilities: pm bus_master cap_list rom ethernet physical tp mii fibre 10bt 10bt-fd 100bt 100bt-fd autonegotiation
       configuration: autonegotiation=on broadcast=yes driver=natsemi driverversion=2.1 duplex=full ip=192.168.0.6 latency=32 link=yes maxlatency=52 mingnt=11 multicast=yes port=twisted pair speed=100Mbit/s
       resources: irq:21 ioport:1000(size=256) memory:e3104000-e3104fff memory:f9200000-f920ffff
----------------
Nowt wrong with the network at any other level. Blindingly fast local transfers at full 100Mbps, decent internet with no issues at all. Unix networking was my stock in trade for many tears..trust me this isnt a noob misconfiguration.

What is more interesting is netstat

Code: Select all

sudo netstat -a | grep ESTABLISHED
tcp        0      0 192.168.0.6:54187       four.schnuerpel.eu:nntp ESTABLISHED
tcp        0      0 192.168.0.6:954         tempest:nfs             ESTABLISHED
tcp        0      0 192.168.0.6:54186       four.schnuerpel.eu:nntp ESTABLISHED
tcp6       0      0 ip6-localhost:51474     ip6-localhost:ipp       ESTABLISHED
tcp6       0      0 ip6-localhost:ipp       ip6-localhost:51474     ESTABLISHED
I am not running the ssh client at the moment
But you can see two permanent nntp connections one of which is being polled for new headers, and the other of which is dormant

Now if I try and read news

Image
is what happens..

Code: Select all

tcp        0      0 192.168.0.6:36865       ea-in-f136.1e100.:https ESTABLISHED
tcp        0      0 192.168.0.6:802         vps.templar.co.uk:nfs   ESTABLISHED
tcp        0      0 192.168.0.6:43998       ea-in-f101.1e100.:https ESTABLISHED
tcp        0      0 192.168.0.6:54189       four.schnuerpel.eu:nntp ESTABLISHED
tcp        0      0 192.168.0.6:954         tempest:nfs             ESTABLISHED
tcp6       0      0 ip6-localhost:51474     ip6-localhost:ipp       ESTABLISHED
tcp6       0      0 ip6-localhost:ipp       ip6-localhost:51474     ESTABLISHED
And you will note that one NNTP session has vanished

(never mind the extra NFS that's has popped up. that's down to me uploading a picture)

and then after successfully downloading a post by trying again

Code: Select all

sudo netstat -a | grep ESTABLISHED
tcp        0      0 192.168.0.6:54199       four.schnuerpel.eu:nntp ESTABLISHED
tcp        0      0 192.168.0.6:802         vps.templar.co.uk:nfs   ESTABLISHED
tcp        0      0 192.168.0.6:954         tempest:nfs             ESTABLISHED
tcp6       0      0 ip6-localhost:51474     ip6-localhost:ipp       ESTABLISHED
tcp6       0      0 ip6-localhost:ipp       ip6-localhost:51474     ESTABLISHED
then after a few minutes, up pops another nntp session

Code: Select all

tcp        0      0 192.168.0.6:54199       four.schnuerpel.eu:nntp ESTABLISHED
tcp        0      0 192.168.0.6:54200       four.schnuerpel.eu:nntp ESTABLISHED
tcp        0      0 192.168.0.6:802         vps.templar.co.uk:nfs   ESTABLISHED
tcp        0      0 192.168.0.6:954         tempest:nfs             ESTABLISHED
tcp6       0      0 ip6-localhost:51474     ip6-localhost:ipp       ESTABLISHED
tcp6       0      0 ip6-localhost:ipp       ip6-localhost:51474     ESTABLISHED
and its still working, for now..

Now I dont pretend to understand how thunderbird manages its connections to news servers, but whenever there are two nntp sessions left for any length of time, that error message pops up and they get replaced with one, then two new ones.

I cant show you anything meaningful on netsat/ssh because having put in keepalives, it behaves flawlessly.

What may be worth trying is to see whether or not the issue happens with local non-NATTTED sessions.

Code: Select all

sudo netstat -a | grep ESTABLISHED
tcp        0      0 192.168.0.6:54220       four.schnuerpel.eu:nntp ESTABLISHED
tcp        0      0 192.168.0.6:954         tempest:nfs             ESTABLISHED
tcp        0      0 192.168.0.6:45242       tempest:telnet          ESTABLISHED
tcp        0      0 192.168.0.6:54221       four.schnuerpel.eu:nntp ESTABLISHED
tcp6       0      0 ip6-localhost:51474     ip6-localhost:ipp       ESTABLISHED
tcp6       0      0 ip6-localhost:ipp       ip6-localhost:51474     ESTABLISHED
trying a local telnet session.

However I think telnet has its own keepalives so it may not actually prove anything. Except that if it too times out, the NAT router is not to blame.

Hmm. I'll report back on the telnet, and try rebooting the router just in case.

be right back after decent pause, and coffee.

leo.smith
Level 1
Level 1
Posts: 24
Joined: Wed Jan 02, 2013 12:23 pm

Re: TCP sessions timing out?

Post by leo.smith » Thu Sep 18, 2014 10:47 am

Curiouser and curiouser, when not reading news, the two nntp sessions periodically get replaced

Vioz:

Code: Select all

sudo netstat -a | grep -E 'nntp|telnet'
tcp        0      0 *:telnet                *:*                     LISTEN     
tcp        0      0 192.168.0.6:45242       tempest:telnet          ESTABLISHED
tcp        0      0 192.168.0.6:54263       four.schnuerpel.eu:nntp ESTABLISHED
tcp        0      0 192.168.0.6:54264       four.schnuerpel.eu:nntp ESTABLISHED

sudo netstat -a | grep -E 'nntp|telnet'
tcp        0      0 *:telnet                *:*                     LISTEN     
tcp        0      0 192.168.0.6:54298       four.schnuerpel.eu:nntp ESTABLISHED
tcp        0      0 192.168.0.6:54299       four.schnuerpel.eu:nntp ESTABLISHED
tcp        0      0 192.168.0.6:45242       tempest:telnet          ESTABLISHED
Note that the source port numbers have changed...indicating both sessions were torn down and new ones added.

I am now thoroughly confused. This one may be nothing to do with networking at the base level, but due to perhaps an interaction between 'new thunderbird' and 'old albasani news server'

User avatar
PatH57
Level 13
Level 13
Posts: 4576
Joined: Tue Mar 25, 2014 12:11 pm
Location: here and there

Re: TCP sessions timing out?

Post by PatH57 » Thu Sep 18, 2014 11:02 am

just tried the news reader and got a connection refused? a minute later it worked.
Also in some rare occasions it seems that the kernel 3.13 does drop connections (could explain why you have sometimes 2 on a very near port mapping)
Try a 3.14 kernel or wait to see if it's not related to the news server?
People disagree with me. I just ignore them.
(Linus Torvalds, regarding the use of C++ for the Linux kernel.)

Please Add [Solved] to the topic-title of your first post when appropriate so others know they might find a solution here.

leo.smith
Level 1
Level 1
Posts: 24
Joined: Wed Jan 02, 2013 12:23 pm

Re: TCP sessions timing out?

Post by leo.smith » Thu Sep 18, 2014 11:13 am

I am loathe to go 3.14 if its still unstable..

But I guess I can always select 3.13 at boot time.

Mmmm. I see listed many flavours of 3.13 but none of 3.14

What is the 'approved' way to get a later 3.13 kernel, and 3.14, which aint in the repo
??

Ok update manager takes me up to 3.13.35, so I'll try that first

I don't know why thunderbird works that way, I think it opens one for headers and the other for post bodies. But that is pure guesswork.

Later, but not 3.14, kernel installed. Will reboot now.

Hope it works ;-)

...it did, but made no difference. that's 3.13.35 flavour.

User avatar
PatH57
Level 13
Level 13
Posts: 4576
Joined: Tue Mar 25, 2014 12:11 pm
Location: here and there

Re: TCP sessions timing out?

Post by PatH57 » Thu Sep 18, 2014 11:44 am

I always use this way to do it (just have to change the version)
http://dcageek.blogspot.fr/2014/06/inst ... buntu.html
People disagree with me. I just ignore them.
(Linus Torvalds, regarding the use of C++ for the Linux kernel.)

Please Add [Solved] to the topic-title of your first post when appropriate so others know they might find a solution here.

leo.smith
Level 1
Level 1
Posts: 24
Joined: Wed Jan 02, 2013 12:23 pm

Re: TCP sessions timing out?

Post by leo.smith » Thu Sep 18, 2014 11:50 am

PatH57 wrote:I always use this way to do it (just have to change the version)
http://dcageek.blogspot.fr/2014/06/inst ... buntu.html
Thanks for that Pat, but I think I am going to pass. I am thinking this may well be another issue than the kernel.

No doubt 3.14 will turn up as an upgrade in mint 17 standard distro in due course.

I hope so.

this is really a minor irritation and I just hoped someone had a guaranteed fix.
I suspect its something between T'bird and albasani not playing nice.

User avatar
Lanser
Level 4
Level 4
Posts: 298
Joined: Mon Mar 08, 2010 5:12 am
Location: Salzburg Austria

Re: TCP sessions timing out?

Post by Lanser » Sat Sep 27, 2014 9:50 pm

Hi Leo. I am seeing very similar issues on my LAN with LM17 clients.
But I had put the problem down to a potentially defective L2 Smart Switch dropping packets, port to port, within a tagged VLAN.
Just curious if you are using smart or passive switches ?

Lanser
Thinkpads:- LM19.x Mate, LMDE3, Debian 10

Post Reply

Return to “Other networking topics”