mcumming
Unregistered
|
|
I've been running Webload tests where I get log window "Server not responding" warnings although my test server appears to have plenty of available CPU time and memory and appear to be responding quite quickly to the majority of the requests being made.
Only 1 or 2 http requests out of every 100 or so return this warning and they happen somewhat sporadically during my test run.Has anyone else seen this? I've even tried manually logging into my test server during test runs just as a sanity check and the reponse time is great.
Also, I've searched through the documentation and on Radview's site - does anyone know of a trigger type function you can call when a thread receives a warning message?
I don't mean the WarningMessage() function, which lets you raise a custom warning message, I need something that can tell me when one of my test threads gets a "server not responding" (or any other) warning so that I can handle that condition.
Something analagous to OnScriptAbort() but for warning messages.
Thanks,
Matt
------------------
|
QAchick
Junior Member
Reged: 11/19/01
Posts: 21
|
|
The server not responding is a server error that comes from the server. Webload is just passing the information alongand it is not generated within webload. There is no way that I know of to script around the message coming from the server so you can put in your own functionality. Have you checked your server logs to see what is happening when your test chokes with webload? I would put in some InfoMessages() into your script to see where it is choking and proceed from there.------------------
|
mcumming
Unregistered
|
|
I have lots of InfoMessage() calls in my scripts but I didn't realize that that message was coming from my server - it could just be a bug/bottleneck in my app, I'm still working on determining what exactly is happening.
Thanks for the tip,
Matt------------------
|
cjeans
Junior Member
Reged: 03/27/02
Posts: 6
Loc: Atlanta, GA
|
|
I think the "server not responding" is basically a server timeout. How is the server doing? Even though your WebLoad script seems to work most of the time and your manual browser test is ok, your server could still be having problems. I have seen response times and server not responding messages because of periodic queuing or very high context switching on the server.
|
mcumming
Unregistered
|
|
Thanks for the input cjeans, I'll check those metrics in the next set of tests that I run.
I'm just wondering why I can never reproduce this manually (with a browser) while a test is running. ------------------
|
cjeans
Junior Member
Reged: 03/27/02
Posts: 6
Loc: Atlanta, GA
|
|
I think you would be able to reproduce it manually with a browser if you happen to make one of those 1 or 2 requests of 100. Keep trying!!If you get the error manually, make sure to play the lottery as well! Regards!
|
Phil Hollows
Member
Reged: 06/26/00
Posts: 133
Loc: MA, USA
|
|
Just because your server CPU isn't being stressed too much doesn't mean it can handle all the incoming requests. For example, some web servers have a finite pool of threads that can handle requests. Other inbound requests will have to wait until a thread in the pool frees up. Servicing requests may not be too demanding on the machine but still take time. If the pool size is, say, 10 (and some servers have very low default pool sizes), then by the time the server gets around to servicing the 100th thread that thread might have timed out.You'll probably find that tweaking the web server's configuration will help (check the size of socket, thread, process and other shared resource pools -- if you make a change you'll probably have to restart the server). If the host server were stressed then this tactic wouldn't work (for what I hope are obvious reasons). If there's an application or database server around make similar checks on these systems too. One other thing to check: it might be possible that you're getting a timeout becuase the WebLoad load generator is too stressed. Just make sure that when you're running the load session that your load generator itself ahs enough CPU headroom and is not using more than its physical (not virtual / swap) RAM. Phil ------------------
[This message has been edited by Phil Hollows (edited 04-03-2002).]
|
cjeans
Junior Member
Reged: 03/27/02
Posts: 6
Loc: Atlanta, GA
|
|
Great comments Phil!I actually got that message because we were pushing our T1 too hard. So make sure your network can handle the load as well.
|
Phil Hollows
Member
Reged: 06/26/00
Posts: 133
Loc: MA, USA
|
|
Good point on availble network bandwidth -- thanks. I think to be fair the "Server not responding" message is coming from WebLoad, but what it's doing is reflecting a low level socket timeout error message from the OS. It's certainly not a message the server is somehow sending to the client -- it indicates that the client has been unable to communicate correctly with the target server in the time frame. It can be confusing, I agree.Also worth pointing out that cjeans and my comments on bottlenecks are independent of WebLoad - they would apply to any other load tool encountering this situation too. FWIW Phil
|
Ziggy1
Newbie
Reged: 01/23/08
Posts: 3
|
|
I've been having this message, too.
Instead of coming from the destination server, it appears to be an issue with the proxy server (TestTalk) and relates somehow to "Authentication failing".
The proxy server does not pass the default authentication that Internet Explorer passes when challenged.
Unfortunately, there are three separate fields for authentication (all settable from same dialog box):
Under Global Authentication Settings:
(1) UserName
(2) NTUserName
Under Proxy Authentication Settings:
(3) ProxyUserName
Unfortunately, that leaves 7-permutations (using the above three fields) to try to figure out which if any of these need to be set properly. I created a ServiceUser account on the target machine to be authenticated -- and have tried all 7 possible permutations of applying that User to the fields. No luck.
Funny thing, this used to work.
Moan!
|
Ziggy1
Newbie
Reged: 01/23/08
Posts: 3
|
|
Update! 
RadView appears to be doing a bait-and-switch with their concept of "open source" for WebLoad. (http://www.testingreflections.com/node/view/5922)
The NTLM network authentication module is crippled in the OpenSource flavor (so RadView can upsell their working flavor -- their Professional flavor of WebLoad) -- so you'll just spin your wheels thinking it will work (based upon the dialog boxes.) (Although, I swear it WAS working for me at one time, so it may be enabled intermittantly in their Open Source flavor.)
Instead... expect to see "SERVER NOT RESPONDING" errors to disguise their deprecated OpenSource flavoring!
I'm beginning to wonder if their strategy is to simply get the openSource developers to kick-in code for free so they can use it in their profession version, smack against the concept of an OpenSource license.
|
Previous
Index
Next
Threaded
Edit
Reply
Quote
