At 5 a.m. my internet connection went 'weird'



Traceroute revealed that some internet sites were inaccessible whilst
others going via the same router were just fine.

I reported it to my ISP, sincere problem seemed to be in one of their
boundary routers, and patched into a proxy on a machine I COULD reach,
and (most) traffic was restored.

At 3pm the problem still existed, so I reported it to them by email

At 4 pm I lost patience and phoned. They admitted the problem, but said
that the router in question had been rebooted and the problem had gone,
but 'the solution seems to be to restart the customer router'

I did, and full connectivity was restored.

But the puzzle remains: how could a problem on a machine miles away
reflect itself in corrupted router NAT tables? As this is the only thing
that I can see that could have produced this behaviour.

I am honestly stumped on this.

Anyone provide a rationale?



--
Ineptocracy

(in-ep-toc-ra-cy) €“ a system of government where the least capable to
lead are elected by the least capable of producing, and where the
members of society least likely to sustain themselves or succeed, are
rewarded with goods and services paid for by the confiscated wealth of a
diminishing number of producers.

On 08/01/2013 21:03, The Natural Philosopher wrote:
On 08/01/13 20:42, alexd wrote:
The Natural Philosopher (for it is he) wrote:

Anyone provide a rationale?

Perhaps the router [or datacentre] that your session terminated on had a
problem and rebooting the router got you on to a different router/DC.

Nope. traceroute reveals an identical route. IP is static

I have got as far as conjecturing that the NAT tables for even broken
routes are preserved after dropping the TCP connection.

So that e.g I would be using the same source port every time.

OTOH its possible that routes BACK to me - OSPF or similart - were
broken in some way and reconnecting deleted those.

BUT those operate on IP address alone, not destination (my source) port,
as such. So that falls apart.


Not sure who you are with but +net have been having quite a few problems
today with routers etc and late afternoon here, their DNS server was
playing up which very much confused us as the lappie was Ok but is
actually using a fixed DNS server not theirs.
Lots of switching of/on of router etc until realised.
Always teh obvious.
Thats life (well internet life anyway)

Mike

On 08/01/2013 22:33, MikeJ wrote:
On 08/01/2013 21:03, The Natural Philosopher wrote:
On 08/01/13 20:42, alexd wrote:
The Natural Philosopher (for it is he) wrote:

Anyone provide a rationale?

Perhaps the router [or datacentre] that your session terminated on had a
problem and rebooting the router got you on to a different router/DC.

Nope. traceroute reveals an identical route. IP is static

I have got as far as conjecturing that the NAT tables for even broken
routes are preserved after dropping the TCP connection.

So that e.g I would be using the same source port every time.

OTOH its possible that routes BACK to me - OSPF or similart - were
broken in some way and reconnecting deleted those.

BUT those operate on IP address alone, not destination (my source) port,
as such. So that falls apart.


Not sure who you are with but +net have been having quite a few problems
today with routers etc and late afternoon here, their DNS server was
playing up which very much confused us as the lappie was Ok but is
actually using a fixed DNS server not theirs.
Lots of switching of/on of router etc until realised.
Always teh obvious.
Thats life (well internet life anyway)

Its been wider than Plusnet - I had quite a few calls this afternoon
with various people loosing broadband or getting erratic connectivity.
Interestingly even sites with redundant links via separate ISPs were
getting the same problems on both WAN connections.

--
Cheers,

John.

/================================================== ===============\
| Internode Ltd - http://www.internode.co.uk |
|-----------------------------------------------------------------|
| John Rumm - john(at)internode(dot)co(dot)uk |
\================================================= ================/

On 08/01/13 23:20, John Rumm wrote:
On 08/01/2013 22:33, MikeJ wrote:
On 08/01/2013 21:03, The Natural Philosopher wrote:
On 08/01/13 20:42, alexd wrote:
The Natural Philosopher (for it is he) wrote:

Anyone provide a rationale?

Perhaps the router [or datacentre] that your session terminated on
had a
problem and rebooting the router got you on to a different router/DC.

Nope. traceroute reveals an identical route. IP is static

I have got as far as conjecturing that the NAT tables for even broken
routes are preserved after dropping the TCP connection.

So that e.g I would be using the same source port every time.

OTOH its possible that routes BACK to me - OSPF or similart - were
broken in some way and reconnecting deleted those.

BUT those operate on IP address alone, not destination (my source) port,
as such. So that falls apart.


Not sure who you are with but +net have been having quite a few problems
today with routers etc and late afternoon here, their DNS server was
playing up which very much confused us as the lappie was Ok but is
actually using a fixed DNS server not theirs.
Lots of switching of/on of router etc until realised.
Always teh obvious.
Thats life (well internet life anyway)

Its been wider than Plusnet - I had quite a few calls this afternoon
with various people loosing broadband or getting erratic connectivity.
Interestingly even sites with redundant links via separate ISPs were
getting the same problems on both WAN connections.

I'll peruse the Register and see if there's a hint of what's up.

Its probably deployment and testing of the latest deep pan PIzza
inspection, with added cheese and anchovies causing net burps.



--
Ineptocracy

(in-ep-toc-ra-cy) €“ a system of government where the least capable to
lead are elected by the least capable of producing, and where the
members of society least likely to sustain themselves or succeed, are
rewarded with goods and services paid for by the confiscated wealth of a
diminishing number of producers.

On Tuesday, January 8, 2013 11:20:19 PM UTC, John Rumm wrote:
On 08/01/2013 22:33, MikeJ wrote:

On 08/01/2013 21:03, The Natural Philosopher wrote:

On 08/01/13 20:42, alexd wrote:

The Natural Philosopher (for it is he) wrote:



Anyone provide a rationale?



Perhaps the router [or datacentre] that your session terminated on had a

problem and rebooting the router got you on to a different router/DC.



Nope. traceroute reveals an identical route. IP is static



I have got as far as conjecturing that the NAT tables for even broken

routes are preserved after dropping the TCP connection.



So that e.g I would be using the same source port every time.



OTOH its possible that routes BACK to me - OSPF or similart - were

broken in some way and reconnecting deleted those.



BUT those operate on IP address alone, not destination (my source) port,

as such. So that falls apart.





Not sure who you are with but +net have been having quite a few problems

today with routers etc and late afternoon here, their DNS server was

playing up which very much confused us as the lappie was Ok but is

actually using a fixed DNS server not theirs.

Lots of switching of/on of router etc until realised.

Always teh obvious.

Thats life (well internet life anyway)



Its been wider than Plusnet - I had quite a few calls this afternoon

with various people loosing broadband or getting erratic connectivity.

Interestingly even sites with redundant links via separate ISPs were

getting the same problems on both WAN connections.



My ADSL connection went down about 6 minutes past midnight today, then failed CHAP authentication for about 15 minutes before everything was OK again. ISP is entanet. No idea if there is any connection (no pun intended).
Simon.

On 08/01/2013 21:03, The Natural Philosopher wrote:
On 08/01/13 20:42, alexd wrote:
The Natural Philosopher (for it is he) wrote:

Anyone provide a rationale?

Perhaps the router [or datacentre] that your session terminated on had a
problem and rebooting the router got you on to a different router/DC.

Nope. traceroute reveals an identical route. IP is static

I have got as far as conjecturing that the NAT tables for even broken
routes are preserved after dropping the TCP connection.

So that e.g I would be using the same source port every time.

OTOH its possible that routes BACK to me - OSPF or similart - were
broken in some way and reconnecting deleted those.

BUT those operate on IP address alone, not destination (my source) port,
as such. So that falls apart.

Plusnet by any chance? If so, BGP was flapping on one of the routers in
our core network. The reboot of your router was probably a coincidence.

If it wasn't Plusnet then ignore me

--
|Bob Pullen Broadband Solutions for
|Support Home & Business @
|Plusnet Plc. www.plus.net
+--------------- twitter.com/plusnet ----------------

On 09/01/13 11:44, Plusnet Support Team wrote:
On 08/01/2013 21:03, The Natural Philosopher wrote:
On 08/01/13 20:42, alexd wrote:
The Natural Philosopher (for it is he) wrote:

Anyone provide a rationale?

Perhaps the router [or datacentre] that your session terminated on had a
problem and rebooting the router got you on to a different router/DC.

Nope. traceroute reveals an identical route. IP is static

I have got as far as conjecturing that the NAT tables for even broken
routes are preserved after dropping the TCP connection.

So that e.g I would be using the same source port every time.

OTOH its possible that routes BACK to me - OSPF or similart - were
broken in some way and reconnecting deleted those.

BUT those operate on IP address alone, not destination (my source) port,
as such. So that falls apart.

Plusnet by any chance? If so, BGP was flapping on one of the routers in
our core network. The reboot of your router was probably a coincidence.

If it wasn't Plusnet then ignore me

No, it was in fact IDnet.
But it confirms one suspicion, that there was something odd going on in
the peering between major ISPs.


--
Ineptocracy

(in-ep-toc-ra-cy) €“ a system of government where the least capable to
lead are elected by the least capable of producing, and where the
members of society least likely to sustain themselves or succeed, are
rewarded with goods and services paid for by the confiscated wealth of a
diminishing number of producers.

On 09/01/2013 12:42, The Natural Philosopher wrote:
On 09/01/13 11:44, Plusnet Support Team wrote:
On 08/01/2013 21:03, The Natural Philosopher wrote:
On 08/01/13 20:42, alexd wrote:
The Natural Philosopher (for it is he) wrote:

Anyone provide a rationale?

Perhaps the router [or datacentre] that your session terminated on
had a
problem and rebooting the router got you on to a different router/DC.

Nope. traceroute reveals an identical route. IP is static

I have got as far as conjecturing that the NAT tables for even broken
routes are preserved after dropping the TCP connection.

So that e.g I would be using the same source port every time.

OTOH its possible that routes BACK to me - OSPF or similart - were
broken in some way and reconnecting deleted those.

BUT those operate on IP address alone, not destination (my source) port,
as such. So that falls apart.

Plusnet by any chance? If so, BGP was flapping on one of the routers in
our core network. The reboot of your router was probably a coincidence.

If it wasn't Plusnet then ignore me

No, it was in fact IDnet.

Interesting - my IDnet line was also a touch "variable" last night.
Which along with bits of plusnet not behaving as normal rendered my dual
redundant connections somewhat dually redundant at times!

Clients on Plusnet and IDnet combinations seem ok today. Only one
customer still having grief, but that looks like openreach are the
problem since both their lines are disconnecting randomly.

But it confirms one suspicion, that there was something odd going on in
the peering between major ISPs.



--
Cheers,

John.

/================================================== ===============\
| Internode Ltd - http://www.internode.co.uk |
|-----------------------------------------------------------------|
| John Rumm - john(at)internode(dot)co(dot)uk |
\================================================= ================/