[Dxspider-support] Spider stall.

[Kjell Jarl]

Example netstat -tupno.
The unkn-4 seems to appear to a windows machine.
It is the sendqueue that fills up, it seems.
Th rest of connections are having received queue, as the last example.


Proto Recv-Q Send-Q Local Address           Foreign Address 
State       PID/Program name    Timer
tcp      125  13160 85.24.164.91:4168       146.83.196.31:9000 
ESTABLISHED 813/perl            unkn-4 (69.64/0/1)
tcp    79177      0 85.24.164.91:4160       212.51.10.184:8000 
ESTABLISHED 813/perl            keepalive (2631.88/0/0)

I can send a larger file off the list.
73
Kjell


Mike McCarthy, W1NR wrote:
> I have a theory...
> 
> Since Spider is a single threaded application, any system call that blocks
> will cause the entire process to appear to hang.  It is most likely a TCP
> call that is hanging.  Once I observed that one user appeared to be
> connected but was in fact trying to reconnect with no success.  There were a
> lot of TCP sockets open in the SYN state from him trying to reconnect.  His
> IP address still had a socket in the connected state with a ton of traffic
> queued up waiting for him.
> 
> The only way I know of to try to "fix" this is to have each connection in a
> separate process or thread from the main application.  If one does get hung,
> it doesn't prevent the rest of the system from functioning.  I don't know if
> that is possible in Perl.  Again, this is just a theory.  It would help to
> see the output of netstat when this occurs.
> 
> AX25d does spawn "client" for each RF connection and I don't see it ever
> hang despite numerous filled pipes from marginal or RF users who switch off
> the radio without a proper disconnect.  Perhaps the way to fix this is to
> spawn client via xinetd for each telnet connect?
> 
> I have seen this type of hang a couple of times and it always clears itself
> after about 20 minutes.  Once, and only once, I had a system hang so bad I
> could not get a response from the console.  But this was a couple of years
> and several versions of Linux ago and may have had nothing to do with
> spider.
> 
> Mike, W1NR
> 
> -----Original Message-----
> From: dxspider-support-bounces at dxcluster.org
> [mailto:dxspider-support-bounces at dxcluster.org] On Behalf Of Kjell Jarl
> Sent: Sunday, October 08, 2006 8:36 AM
> To: rene_olsen at post3.tele.dk; The DXSpider Support list
> Subject: Re: [Dxspider-support] Spider stall.
> 
> Hi,
> What OS?
> I have seen this many times on Linux 2.4. I suspect that TCP stack is filled
> due to a bad TCP/IP link and spider can not deliver data.
> 73
> Kjell
> 
> 
> Rene Olsen wrote:
> 
>>Hi.
>>
>>Today the spider at oz5bbs-7 decided to suddenly stall. I have seen it 
>>on one other occation, and there is absolutely no indication on the debug
> 
> log, as to what is going on.
> 
>>Running spider 1.52 build 61.521 (should be latest version) and perl
> 
> 5.6.1.
> 
>>The debug log looks like this:
>>
>>1160298101^<- I SK3W-6 PC11^7060.0^IZ8DDP/P^08-Oct-2006^0904Z^DCI SA- 
>>062^I1SCL^IR4AD^H95^~
>>1160298101^PCPROT: Duplicate Spot ignored
>>1160298102^<- I N0VD PC11^7060.0^IZ8DDP/P^08-Oct-2006^0904Z^DCI SA- 
>>062^I1SCL^IR4AD^H94^~
>>1160298102^PCPROT: Duplicate Spot ignored
>>1160298102^<- I OZ2DXC PC11^7060.0^IZ8DDP/P^08-Oct-2006^0904Z^DCI SA- 
>>062^I1SCL^IR4AD^H92^~
>>1160298102^PCPROT: Duplicate Spot ignored
>>1160298102^<- I DB0SUE-7 PC11^7060.0^IZ8DDP/P^08-Oct-2006^0904Z^DCI 
>>SA- 062^I1SCL^IR4AD^H4^~
>>1160298102^PCPROT: Duplicate Spot ignored 1160298102^-> D EI7SDX 
>>PC51^EI7SDX^OZ5BBS-7^1^ 1160302409^DXSpider V1.52, build 61.521 
>>started
>>
>>I looked at least 200 lines back in the debug log, and don't see 
>>anything that could indicate the reason.
>>
>>>From the log I can see that it was dead for almost 72 minutes, and 
>>
>>>there were nothing else
>>
>>to do than to kill the proccess and start it again. Couldn't log in 
>>and shut down the normal way.
>>
>>Any idea as to what this can be. It is the second time that this 
>>happens within a month I think.
>>
>>Vy 73 de Rene / OZ1LQH
>>
>>
>>
>>_______________________________________________
>>Dxspider-support mailing list
>>Dxspider-support at dxcluster.org
>>http://mailman.tobit.co.uk/mailman/listinfo/dxspider-support
>>
>>
> 
> 
> 
> 
> _______________________________________________
> Dxspider-support mailing list
> Dxspider-support at dxcluster.org
> http://mailman.tobit.co.uk/mailman/listinfo/dxspider-support
> 
> 
> 
> _______________________________________________
> Dxspider-support mailing list
> Dxspider-support at dxcluster.org
> http://mailman.tobit.co.uk/mailman/listinfo/dxspider-support
> 
>