httpd-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Bill Moran <>
Subject Re: [users@httpd] Slow/No response from Apache 1.3.27 on a Netware 6.0 Server
Date Wed, 10 Sep 2003 14:38:00 GMT
Michael McKenzie wrote:
> Software Involved:  Netware 6.0, GroupWise WebAccess 6.5, Tomcat 33, 
> Apache 1.3.27
> Hardware:  HP Netserver (P3 with 2G RAM, 70+ MB HD, Onboard NIC with 
> CE100 Drivers)
> Network:  Connected to a Cisco Switch
> I’ve installed a server that is functioning as a Groupwise WebAccess 
> server only.  There are no other apps running on it.  Essentially, this 
> server takes requests from users through a browser and passes those 
> requests to another Netware server that is running the GroupWise App.  
> At random intervals, the WebAccess server stops responding to new 
> requests from people who are trying to login to the web client (or 
> becomes very slow—upwards of 2 minutes to load a simple 2 text box form 
> for login). 

That sounds like some sort of timeout.  i.e. DNS, or some other related
service that is not working correctly.

> Some troubleshooting that I’ve done up to this point include seeing 
> which ports and which apps are working for the server.  During a period 
> of unresponsiveness, I am able to connect to a management interface on 
> the server through IE using port 2200 on a https connection.  There 
> appears to be no unresponsiveness on this management interface.  If I 
> try to connect to http://ipaddress <http://ipaddress/>, there is no 
> response.  If I try to connect to http://ipaddress/servlet/webacc, there 
> is no response.  And if I try to connect to 
> https://ipaddress/servlet/webacc, there is no response.  I suppose the 
> main thing that I’m seeing with this is that if tomcat was the issue, 
> then it should only be unresponsive when connecting to the 
> servlet/webacc path.  In this instance, even port 80 doesn’t work.  I 
> can’t understand how apache is providing the connectivity for both 2200 
> & 80, and one works and one doesn’t.  The problem doesn’t appear to be 
> load related.  This can happen on a weekday when 150+ users are 
> accessing it or on a Sunday night when there is virtually no one on the 
> system.

I would suggest you review your config file and establish an understanding
of how Apache is configured to provide services on multiple ports.
Otherwise, you're unlikely to understand enough to diagnose what's going

> Another interesting observation is that when the problem is happening, 
> we can still see users access, sending, reading, mail through the 
> webaccess agent.  This indicates they are communicating properly with 
> the server on http://ipaddress/servlet/webacc.  They usually do report a 
> slow down in response from the server, but they are still able to remain 
> connected and working.

Sounds like some sort of problem with session management.  i.e. the
server is able to manage existing sessions, but unable to create more.

Review the system used for session management for possible problems.

> We’ve looked over the server and see nothing that stands out.  No 
> extraordinary cpu utilization, network utilization or memory 
> utilization.  Everything looks normal.

That would seem to indicate a timeout on some supportive service.  Since
someone, somewhere, is simply waiting for an answer that isn't going to
come, there's not unusual CPU, disk, or network usage.  Using a packet
capture tool, such as Ethereal, could possible reveal what's really
going on.  If a timeout is the problem, you'll see repeated requests
from a particular host (usually at fixed interval) that gets no response.

> We did increase the ThreadPerChild from 50 to 250.  This is even unusual 
> though.  The users are connecting through a Bordermanager 3.6 Proxy 
> server which is forwarding the request to the web server.  Looking at 
> the TCP/IP stats, you can only see around 5 port 80 connections from 
> this server even if there are 150 users online at the time.  I don’t 
> quite understand this.

Because http is stateless, you'll only see connections for users who are
actually requesting a page at the moment you check.  5/150 doesn't sound
unusual for something like that, as clients who are (for example) reading
mail are logged in but not requesting new pages.

One potential debugging step would be to remove the Bordermanager from
the loop and see if the problem goes away.  If you simplify your config,
it'll be easier to find the problem.

> One last shot in the dark…has anyone ever seen DNS play a role in an 
> issue like this???  There isn’t a DNS server on the network, and BM is 
> providing the translation from to the ipaddress.

Yes.  Improperly configured DNS often results in VERY weird responsiveness
for many services.  And by "improperly", you have to consider what the
workstation _expect_ the DNS to look like compared to how you actually
have it set up (i.e. no DNS).
Newer Windows machines expect a working DNS config for all machines they
will be contacting by default.  It's something to verify in any case.

> I’ve really hit a brick wall with this problem and don’t know what else 
> to try, other than building a new server on new hardware and seeing if 
> the same problem still exists.   If anyone can think of anything that we 
> might be able to try, I would greatly appreciate it.

Have you reviewed the Apache access and error logs for clues?  Are there
tomcat or GroupWise logs to review?

Sorry this isn't a more specific response, but with what you've done,
there still seem to be a lot of possible areas where the problem might
be occurring.

Bill Moran
Potential Technologies

The official User-To-User support forum of the Apache HTTP Server Project.
See <URL:> for more info.
To unsubscribe, e-mail:
   "   from the digest:
For additional commands, e-mail:

View raw message