Weird connection problem [ a bit long be patient; plz help ]

Hi,
I've got this weird connection problem with my slack9 server (kernel 2.4.25)
and machines routed through it. There are 2 LANs and 1 DSL connected to this
machine, one LAN uses the DSL (and so it uses this machine as a router) and
the other does not (anyway the 2nd lan is of minor importance to me). The
problem is that from time to time (or should I say, rather often) machines
from LAN that uses DSL get some kind of connection interrupt, it looks like
if someone pulled the cable out of the network card. All possible services
stop working, even ping does not reach the DSL modem connected to server
(connection timeout), but everything works fine from the server side, net is
working all the time, no interrupts/timeouts, the problem exists only on the
routed machines. After about half a minute everything comes back to normal.
I've already checked couple of things: I'm sure that it isn't faulty DSL
modem cause when I've connected the modem directly to my pc everything
worked fine. I've also checked amount of connections and it was range from
30 to 150 so I'm pretty sure that's not the issue either (and conntrack_max
is set to 20480 or so). Also checked network cards (tested about 5 of them
and still no results). Pretty please, help me, I have no idea what's wrong,
and this is a very very annoying problem ! Thx in advance. ** Happy Easter
!!! **
p.s. machine details: p166mmx 64mb 1x intel pro 1x dlink 1x 3com (I know
it's kinda crappy but should be enough to get along with 1 dsl and 5 users
from lan)
--
°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°
krusher : [ GG: 1183577 UIN: 13305122 ]
[ "nick_from_above"@kawiarenka.com.pl ]
°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°

krusher wrote:

> worked fine. I've also checked amount of connections and it was range from
> 30 to 150 so I'm pretty sure that's not the issue either (and conntrack_max
> is set to 20480 or so). Also checked network cards (tested about 5 of them
> and still no results). Pretty please, help me, I have no idea what's wrong,

> p.s. machine details: p166mmx 64mb 1x intel pro 1x dlink 1x 3com (I know
> it's kinda crappy but should be enough to get along with 1 dsl and 5 users
> from lan)

Perhaps You want to clarify one thing first: With Your five LAN users,
dou You get 30 to 150 connections in parallel...? That means that all of
them have 30 connections each. This may be the case, so don't worry.

From what You wrote, there's no way to help You. - I won't send You back
to the "smart questions" site, but Your posting does not provide the
slightest clue as for where to begin.

What exactly do You mean by "cannot ping the DSL modem"? - Can You ping
Your router's external IP? If not, this at least eliminates a possible
DNS problem. - Is there, by chance, any process running on that P166
that regularly uses up much resources?

Do some homework for Yourself first, form a more specific question and
then start a new thread.

Sorry, I'd really like to help, but for that I'd need at least _some_
information from You.


Cheers, Jack.

--
----------------------------------------------------------------------
My personal reading of the string "MicroSoft" expands to "NanoWeak"...

"jack" <(E-Mail Removed)> wrote in message
news:c5a0bf$v5q$01$(E-Mail Removed)...
> Perhaps You want to clarify one thing first: With Your five LAN users,
> dou You get 30 to 150 connections in parallel...? That means that all of
> them have 30 connections each. This may be the case, so don't worry.
>
> What exactly do You mean by "cannot ping the DSL modem"? - Can You ping
> Your router's external IP? If not, this at least eliminates a possible
> DNS problem. - Is there, by chance, any process running on that P166
> that regularly uses up much resources?
>
> Sorry, I'd really like to help, but for that I'd need at least _some_
> information from You.
>
As I said before I've got 2 LANs connected to this p166 one with 5 PCs and
the other one with about a 100 of them, but only those 5 PCs use this server
as a router/gateway to inet constantly, other machines use it from time to
time, only when transfer between two LANs occur. So not all connections are
made by those 5PCs, some are from the second lan (for example ARP, DHCP
requests pings and other - if those kind of connections are counted by
conntrack as connections which I'm not sure of). 2nd thing is that 150
connections was the peak number which I obtained, average is about 45.
"Cannot ping DSL modem" (with my DSL I got 2 ip addresses, one for my use
and the other is the IP of my modem) - when one of those interrupts occured
I checked whether the modem responds to pings from LAN, and it didn't (and
when net is working I get correct responses from DSL modem when pinging it
from my home computer through lan). It looks like if something was blocking
traffic incoming to/through server from the LAN side.
There aren't many services running on this machine, the only thing this p166
should do is routing, and that's what it does. After or during those
interrupts I checked the load with procinfo and it was minor, usually
0.00-0.01 so I don't think that's the case.
One more thing, thos interrupts occur even when I'm alone on this DSL (I've
disconnected all other users to check this).
Sorry for my previous chaotic post, sure hope this one is better, also sorry
for any language mistakes.
cheers
-----
krusher

krusher wrote:
> As I said before I've got 2 LANs connected to this p166 one with 5 PCs and
> the other one with about a 100 of them, but only those 5 PCs use this server
> as a router/gateway to inet constantly, other machines use it from time to
> time, only when transfer between two LANs occur. So not all connections are
> made by those 5PCs, some are from the second lan (for example ARP, DHCP
> requests pings and other - if those kind of connections are counted by
> conntrack as connections which I'm not sure of). 2nd thing is that 150
> connections was the peak number which I obtained, average is about 45.
> "Cannot ping DSL modem" (with my DSL I got 2 ip addresses, one for my use
> and the other is the IP of my modem) - when one of those interrupts occured
> I checked whether the modem responds to pings from LAN, and it didn't (and
> when net is working I get correct responses from DSL modem when pinging it
> from my home computer through lan). It looks like if something was blocking
> traffic incoming to/through server from the LAN side.
> There aren't many services running on this machine, the only thing this p166
> should do is routing, and that's what it does. After or during those
> interrupts I checked the load with procinfo and it was minor, usually
> 0.00-0.01 so I don't think that's the case.
> One more thing, thos interrupts occur even when I'm alone on this DSL (I've
> disconnected all other users to check this).
> Sorry for my previous chaotic post, sure hope this one is better, also sorry
> for any language mistakes.

Well, firstly I don't think that those 150 connections are too much
for the hardware to handle.

One more question: In the event of such interruption, are You able to
start a new ssh or telnet session on the router? - If not, see whether
there is any consuming process running from time to time.

Does Your modem have an IP for itself? - Normally, the second IP is
that of the access concentrator, Your ISP's router. Anyways, are
You absolutely sure that both modem and the link to Your ISPs AC
are functional at all times?

Then, all this may be caused by conflicting interrupts of Your NICs,
or perhaps by faulty hardware.

Sorry, other than this I have no clue...


Cheers, Jack.

--
----------------------------------------------------------------------
My personal reading of the string "MicroSoft" expands to "NanoWeak"...

"jack" <(E-Mail Removed)> wrote in message
news:c5briv$prg$01$(E-Mail Removed)...
[cut]
> One more question: In the event of such interruption, are You able to
> start a new ssh or telnet session on the router? - If not, see whether
> there is any consuming process running from time to time.
>
> Does Your modem have an IP for itself? - Normally, the second IP is
> that of the access concentrator, Your ISP's router. Anyways, are
> You absolutely sure that both modem and the link to Your ISPs AC
> are functional at all times?
>
> Then, all this may be caused by conflicting interrupts of Your NICs,
> or perhaps by faulty hardware.

No, ssh doesn't work, I'm not able to connect to the server in any way
during connection interrupt. Yes, I'm perfectly sure that modem and the link
are 100% OK, cause connected directly to my pc everything works fine. Just
after the interrupt I'm checking load with procinfo, all load meters show
very very low level of cpu usage (0.00-0.01). Yes my modem has its own IP
for software upgrades and configuration purposes from the ISP side I guess.
I don't think IRQ conflict is the issue 'cause every NIC has it's own IRQ,
not used by any other hardware. Maybe faulty NIC, haven't tried that one yet
(exchanged only one of them so far) or maybe mobo .. hope not.
--
krusher

krusher wrote:

> No, ssh doesn't work, I'm not able to connect to the server in any way
> during connection interrupt. Yes, I'm perfectly sure that modem and the link
> are 100% OK, cause connected directly to my pc everything works fine. Just
> after the interrupt I'm checking load with procinfo, all load meters show
> very very low level of cpu usage (0.00-0.01).

Well, because that propcess may be finished by then... - Try to see what
processes are on the list ("ps") and check whether one of them uses lots
of system resources when waking up (bdflush would be one candidate, but
it should not lead to delays of a notable duration). Or, if a major cron
job is being run when those interrupts occur.

How often does this happen at all? And, can You see any regularity as to
_when_ it does?


> Yes my modem has its own IP
> for software upgrades and configuration purposes from the ISP side I guess.
> I don't think IRQ conflict is the issue 'cause every NIC has it's own IRQ,
> not used by any other hardware. Maybe faulty NIC, haven't tried that one yet
> (exchanged only one of them so far) or maybe mobo .. hope not.

No, hopefully not... - How about Your disk space and swap? Under certain
circumstances, low disk space can slow down Your box or even make it
freeze.


Good luck, Jack.

--
----------------------------------------------------------------------
My personal reading of the string "MicroSoft" expands to "NanoWeak"...

"jack" <(E-Mail Removed)> wrote in message
news:c5cc15$cfh$03$(E-Mail Removed)...
> Well, because that propcess may be finished by then... - Try to see what
> processes are on the list ("ps") and check whether one of them uses lots
> of system resources when waking up (bdflush would be one candidate, but
> it should not lead to delays of a notable duration). Or, if a major cron
> job is being run when those interrupts occur.
>
> How often does this happen at all? And, can You see any regularity as to
> _when_ it does?
>
> No, hopefully not... - How about Your disk space and swap? Under certain
> circumstances, low disk space can slow down Your box or even make it
> freeze.

Frankly saying I have never seen any great usage of CPU on this machine.
Running processes: ( keventd, kswapd, bdflush, kupdated, syslogd, klogd,
atd, dhcp, inetd ). About 600mb of hdd space free, about 200mb of swap free.
It happens very randomly, no regularity at all. Sometimes, doesn't happen
for couple of hours and sometimes time after after time. Maybe tommorrow
I'll find some time and juggle a little with NICs.
--
krusher