[Dxspider-support] Followup - Hang hang and some more hang...

[Kelly Jones]

As a followup -

I am able to make this happen on demand.  I have a "load tester" program 
that was written when I was having issue with CLX several years ago (my 
reason for adopting DXSpdier).  If load up about 8 "robot" users in 
addition to the already connected nodes/users, I can make it so nobody else 
can log in.  As soon as I stop the robots, I am able to initiate new logins.

This is obviously a load issue.  Due to the fact I typically have 50-75 
users and 25 node connections, I suspect I see a bit more load/traffic than 
the average cluster.  The unusual thing is that the box CPU is hardly 
breathing, so it's not a maxed out CPU problem.  It's as though something 
within Perl/Spider can't keep up with the amount of data being 
sent/received.  If anybody is game, I can try to "load up" their node with 
"users" and see how your cluster handles things.

I'm not sure where to go from here.  Dirk, ideas?

Kelly - N0VD



At 08:14 AM 11/13/2005, you wrote:
>As I came down to the shack this morning, I was greeted with the same 
>symptoms described below.  In less than 12 hours I was no longer able to 
>log into my node.  As a test, last night I started the node from the 
>command line instead of as a Windows service.  As soon as I hit cntrl-c, 
>my echo was printed on the screen, then the node proceded to shut down - I 
>found this interesting.  It's as though some buffer somewhere fills up 
>then just stops responding.
>
>I received a private email from another sysop that said he started seeing 
>the exact same symptoms during the CQWW SSB contest.  Something *is* going 
>on here, but I don't know what it is.  I do know I can't keep recycling my 
>node every 12 hours.....
>
>Anybody else seeing this?
>
>Kelly
>
>
>
>At 06:21 PM 11/12/2005, you wrote:
>>This is starting to irritate me....
>>
>>It seems that once I get above 40 or so users something stops 
>>working.  The scenario is that I connect to the node, it asks for my 
>>call, I type it in, hit enter... nothing.  It just sits there for the 60 
>>seconds then disconnects me.
>>
>>If I am a user that is already connected, the system is as snappy as can 
>>be.  It only happens to new, incoming connections.  At this moment I have 
>>41 users and 19 nodes.  If I try to connect either as a node or user, I 
>>get the above scenario.
>>
>>This started happening a couple of weeks ago while still running an older 
>>version of Spider on Linux.  I was hoping that it would go away after I 
>>switched to Windows with the latest and greatest of Spider and Activesate 
>>Perl, but apparently not.  Has anybody seen this before or have a 
>>suggestion as to where to look for the problem?  Something that is 
>>interesting is that when it happens, I no longer get an echo back from 
>>the system.  When it's working, the echo returns every keystroke.  The 
>>only way I have found to "fix" it is to recycle the cluster and wait for 
>>it to happens again...
>>
>>Kelly - N0VD
>>
>>
>>
>>
>>_______________________________________________
>>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