Topic Options
#1313 - 01/05/05 09:19 AM 2.59.1 beta 1 notes
jbrunette Offline

Registered: 08/30/99
Posts: 35
[ edit: originally said "Stop"/"Reset and Restart" didn't do anything ]

I ran two different instances of 2.59.1 beta 1, one after the other, and both stopped tracing after about 3 hours. It didn't say "Querying" at the bottom left, it simply acted like I had clicked "stop" but I hadn't. Clicking "Stop" causes it to change to "Resume" and clicking resume causes the graphs at the bottom to advance (depending on how long it's been since it stopped) but it doesn't actually resume tracing. "Reset and Restart" clears the trace and it gets to "Displayed samples 1 to 2" and "Trace count 2" but there isn't actually any graphs displayed anymore. Picking a different host to trace doesn't work either.

Also, I noticed this a few versions ago and it continues to happen with this beta. On my (admittedly old K6 400Mhz, Windows XP) system it takes about 13 seconds for PP to exit after closing the window. The process uses "100%" of the processor during this time until it finally exits. It does this if I have started a trace or if I just opened it and closed it immediately. Earlier versions did not exhibit this behavior.

Edited by jbrunette (01/05/05 10:38 AM)

#1314 - 01/05/05 01:49 PM Re: 2.59.1 beta 1 notes [Re: jbrunette]
Pete Ness Offline

Registered: 08/30/99
Posts: 1106
Loc: Boise, Idaho
Both of these are interesting symptoms that shouldn't be happening...

Problem 1:

On the "3 hours and stopped", what packet type are you using? You'd probably have reported this, but are you seeing an engine errors listed above the time graph? If not, have you changed the [Advanced] "ShowErrorPanel" setting in your ini file to supress any errors?

When outsanding requests exceeds the listed outstanding requests (Edit -> Options, Packet options, Maximum concurrent requests), then PingPlotter skips the trace (it won't send out any more packets until some have completed). It sounds like there is some sort of condition that is causing packets to *never* complete (which would certainly be a bug). Each of the 3 packet types have separate queus for this - ICMP, UDP and TCP. They also each have slightly different logic for determining that a packet is "done". If you change packet types once this condition has been reached, things should start working again (although this doesn't repair the situation internally, it just starts looking to a new "engine" instance for being overloaded). If you switch back to the original packet type, this will stop working again.

OK, so information I need on this one:

1) What packet type are you using?
2) Are you seeing any engine related errors on the main screen, above the time graph. If not, is your ini file setting supressing those error messages? If you're supressing them, please unsupress them so we can get to the bottom of this.

I've run into the situation using the TCP ping engine only - when I suspend my computer, then resume it, things don't recover correctly (corner cases aren't always fully flushed out on beta versions, unfortunately).

Problem 2:

Do you have any idea what version first exhibited the "13 seconds on close" problem? If you have another PC available to try this on, do you see similar symptoms, or is it isolated to this machine specifically?

Just a quick open / close causes this pause to happen? (I'm trying to figure out what logic paths happen every time where a "wait" condition would be introduced that takes CPU). This could be a problem the version check logic - are you set to have PingPlotter automatically check for new versions? If you do a manual version check (Edit -> Options, Internet, Version checking, "Check Now" button), does the status bar show "You have the latest pre-release version of PingPlotter", or does it stay blank?

Thanks (as always) for the reports, Jason!

- Pete

P.S. I'm moving this post to the beta forum.

#1315 - 01/07/05 09:55 AM Re: 2.59.1 beta 1 notes [Re: Pete Ness]
jbrunette Offline

Registered: 08/30/99
Posts: 35
I am currently using "ICMP using Windows DLL". You are correct that switching to a different engine causes the existing trace to resume.

I do get "There are no receivers" errors a couple times a day, although they do not occur (at least, not logged) at the time the tracing stops. For example, the last error occured about an hour before the trace stopped.

Regarding the CPU usage: after some testing, I do not believe it was a particular version of PP, rather network and operating system changes at my office. I checked all the way back to about version 2.30 and they all exhibit the same behavior. Around the time it seemed to start, we had gone from Windows 2000 to Windows XP SP1 and began sitting behind an ISA server. I do have the proxy information set up in PP and I do see "You have the latest pre-release version of PingPlotter" when I click "Check Now".

I checked a Win2000 machine that's not behind the ISA server (same speed as my office PC) and PP exited in about two seconds, taking up all of the CPU during that time. All of the PCs behind our ISA server are WinXP so I can't test one or the other at this time.




Who's Online
0 registered (), 6 Guests and 0 Spiders online.
Key: Admin, Global Mod, Mod