Folding Forum

CBT wrote:

CBT wrote:Dear Uncle,
Would it be possible to make FahMon detect EUE's (esp. for the GPU2-client) and give a visible warning? Maybe in an extra column, so you can show if there have been EUE's and how many. It would probable by usefull to be able to configure the timespan that FahMon verifies: i.e. if you're able to check you system regularly, you only want to be notified of EUE's in the last few hours. Whereas for others it may be usefull to see all EUE's in the last 48h's or week.

TIA,

Corné

I've thought about it a little more and it may be better to add a this feature in the [View]- or [Tools]-menu. Something like 'Parse the logfile for EUE's'. The output could be a table with the columns 'Client-type (CPU, SMP, etc.), Core, Project, date/time'.

T.I.A.,

Corné

www.mtm78.nl/maxfah/fahwatch.exe might help you already

My CPU usage still goes through the roof during automatic reloads with 2.3.4. It regularly hits 99% for a few minutes at a time, then settles back down.

System is XP Pro 32 SP3, 2GB RAM. I pull almost all of my folding data over an SSH tunnel, but I don't think that should matter -- 2.3.2b handled it much better.

I have a similar setup to KE1AH, monitoring one local GPU2 and a remote GPU2 and SMP over an SSH connection (same LAN). Mine's a dual core AMD.2.1 with 2GB, XPP. It's new fresh clean install of 2.3.4 from last night. When it does the auto-reload it drives the processor core it's on to 100% for about 25 seconds, drops back to zero, goes back to 100% for a couple seconds, goes back to zero until next time. Uncle, is that the expected behavior? Between spikes it uses about zero CPU, 11MB Mem and 7MB VM. During the spike it goes to about 100% of it's processor core, 38MB mem and 35MB VM. Seems kinda intensive, no? The only other thing of note running is FF, using about 193MB mem.

It's also reporting the SSH Linux/WINE/GPU2 as hung, doesn't show the WU info but goes right along updating the percent completed as normal. Actually, both of the nSSH Linux clients shoed as hung till I went through with chmod 777 down to the FAHlog.txt. That fixed the LINUX SMP client but not the GPU2.

The short answer is no, that kind of CPU usage is not intended but seems to be common on Windows (I haven't noticed it on Linux). Can you try disabling your networked client in FahMon (right click and select disable) and monitor the CPU usage during updates of the local client only?

finally got back to the computer, long day.

First let me correct what I said last night. I'm using an SSH connection to manage/admin the ubuntu 8.1 folder, but the linux client monitoring is actually done via SMB, not SSH. I've got a drive mapped to a samba share on the linux box, add the clients in fahmon just like they are local. But, you probably already know this.

Now, in direct A-B-A test disabling the remote clients makes the secondary, 1 or 2 second spike go away. The 25 second spike is apparently caused by the Win client monitoring. Seems counter-intuitive since there's only one Win client and it's local but there it is. It is repeatable. Interesting, turning on the kernel time part of the display shows that less than half the spike is kernel time, most of the spike is from userland processes.

see if this image gizmo will work:

nope

how about this link to a pic:

http://picasaweb.google.com/lh/photo/1f ... tdgkF2NqtU

cute, pic disappears. hit the slide show button, it'll show up.

uncle_fungus wrote:The short answer is no, that kind of CPU usage is not intended but seems to be common on Windows (I haven't noticed it on Linux). Can you try disabling your networked client in FahMon (right click and select disable) and monitor the CPU usage during updates of the local client only?

WFO put in a ticket on your site for the same problem, I hope you will be able to fix it

Yep. I'm working on it (I need to do some performance profiling on the Windows build).

uncle_fungus wrote:Yep. I'm working on it (I need to do some performance profiling on the Windows build).

I think the im I send you about the popup box should also be fixed, it also caused fahmon to hang a couple of times and it's not really needed to inform me with a popup box a network client can't be read. Not now you have the ability to disable clients, imo the clients should be automaticly disabeld and the listview row with disabeld clients should be even more noticable ( maybe red color as background? ) and that way people will notice it ). Or check if the clients is minimized in the tray only and show a balloon message on the windows client informing a network client is not accessible instead of the msgbox?

Could you grab a screenshot of the popup box? There's nothing in FahMon that should generate that kind of error, so I suspect it's one generate by Windows.

On the first run of FahMon after the computer has started up I consistently get the following pop up:

Deleted stale lock file '/home/david/.FahMon__LOCK__'.

This is unique to the Linux (Ubuntu) version, I've never seen it on my Windows box. I should also add that it only occurs after startup, if I shut down FahMon and restart it then the message doesn't display.

Oak37 wrote:On the first run of FahMon after the computer has started up I consistently get the following pop up:

Deleted stale lock file '/home/david/.FahMon__LOCK__'.

This is unique to the Linux (Ubuntu) version, I've never seen it on my Windows box. I should also add that it only occurs after startup, if I shut down FahMon and restart it then the message doesn't display.

That's a known "bug". The Linux version of wxWidgets seems incapable of processing the OS_SHUTDOWN or OS_LOGOFF type events which means that when you shutdown or logoff and FahMon is still running, it doesn't close gracefully leading to the error on next startup. The only way around this is to close FahMon before you logoff or shutdown in Linux.

Hi Andrew!

With the current source from the trunk (r386) Visual C++ 2008 gives a linking warning for each .obj file. All warnings are the same:

aboutDialog.obj : warning LNK4254: Abschnitt ".text" (60000020) in ".data" (C0000040) mit verschiedenen Attributen zusammengeführt
base64Codec.obj : warning LNK4254: Abschnitt ".text" (60000020) in ".data" (C0000040) mit verschiedenen Attributen zusammengeführt
benchmark.obj : warning LNK4254: Abschnitt ".text" (60000020) in ".data" (C0000040) mit verschiedenen Attributen zusammengeführt
.
.
.

(The German text means something like: Section ".text" in ".data" linked with different attributes)

However, the compiled fahmon.exe works just fine. I'm just wondering.

Gerby

Those warnings are expected as I'm evaluating some optimisations in the MSVC compiler, one of which combines the text and data segments of the code.

Alright, thanks for this background info. Keep on optimising!

well, since my post of 11/16 the CPU spike has gone from 25 secs to 40 secs and the idle memory usage gone from 12/11 to 21/20 Mem/VMem. maybe a little memory leak? during the refresh it's increasing about 9 - 10MB on both Mem and VMem. Still 2.3.4.