HFM.NET - Monitoring Application for Folding@Home v7

[harlam357]

Looks nice, got to take a closer look at the source though before I'll say more but the layout is very clean.

(Double Post... see below)

[harlam357]

Looks nice, got to take a closer look at the source though before I'll say more but the layout is very clean.

Thanks Mtm. Would love to collaborate with you. In what language is maxFah written? Sorry if it's in there... I just scanned your post. Lots a screen shots!

[harlam357]

New Version: 0.2.0 - Build 20 - Beta - Download

- md5: e00511463fc31a6e75f92cfc4c60419f

- sha1: 939135b95f48392370b4b98bad876974c5f79093

Change Log

• Fix: Only accumulate Completed and Failed counts on most recent unit parse.
• Fix: Final fix for resuming from a long pause. I think I've got it licked now.
• Fix: Parsing of frame percentage from older clients like v5.02.
• Fix: Parsing Frame 0 - which doesn't happen with GPUs and rarely on SMP. However, Standard clients often report Frame 0.
• Fix: Unhandled exception that occurs when viewing the benchmarks window with no clients loaded (actually no client selected).
• Change: Setup frmPreferences to only show the WebBrowser control when not running under Mono Framework.
• Change: Patched the main Project Resources.resx file for case sensitivity.

Upgrading HFM to the latest release

• When upgrading HFM, please take the new Zip file and extract it to the same location as your current install and overwrite the older files.
• The user settings are tied to the location of the HFM.exe file. So if you move or otherwise rename any directory in the path to the HFM.exe file, you will lose your custom settings. Please choose your install location wisely, I recommend a very generic folder name like "\Program Files\HFM". Settings will also not follow the application from machine to machine. It is possible to transfer them and I will post instructions on how to do so at a later time.

Donate to the HFM.NET Project with PayPal

[weedacres]

This 'Offset' is used to determine the 'Status' of the client in question. In the above example you'll see that the Time Of Last Frame is later than Terminal Time. If this is the case, the client is deemed to be running. Terminal Time is calculated differently for GPU clients vs. CPU clients. However, in either case, the Time Of Last Frame must be later than the Terminal Time. If the opposite is true, then the client will be deemed Hung and will not show any progress percentage, TPF, PPD, etc.

If these SMP instances are notfred's VM (or any other VM that reports UTC +0)... yes, you will want to check the 'Client is on Virtual Machine' option. Then Debug with the times as I've stated above.

What I would recommend you do is set the system time on your monitored machines to the correct time. If that is done, you won't have to set any 'Offset' value for your clients. Much simpler approach.

Thanks for the help.
I have 4 Ubuntu SMP's running in VMware. Here's an example of what I'm seeing in the log.

Code: Select all

[6/4/2009-08:16] +  - Retrieval Time (Date) ------- : 6/4/2009 08:16
[6/4/2009-08:16] +  - Time Of Last Frame (TimeSpan) : 15:01:52
[6/4/2009-08:16] +  - Offset (Minutes) ------------ : 0
[6/4/2009-08:16] +  - Time Of Last Frame (Adjusted) : 15:01:52
[6/4/2009-08:16] +  - Prior Day Adjustment -------- : True
[6/4/2009-08:16] +  - Time Of Last Frame (Date) --- : 6/3/2009 15:01
[6/4/2009-08:16] +  - Terminal Time (Date) -------- : 6/4/2009 07:46
[6/4/2009-08:16] + HFM.Instances.ClientInstance.ProcessExisting (Felix 2) Execution Time: 109 ms
[6/4/2009-08:16] - HFM.Instances.ClientInstance.Retrieve (Felix 2) Client Status: Hung

The VM clocks are all within a minute of the host clock and all are sync's from the same time server. I checked to make sure that they are in the correct time zone (PDT). Notice above that the last frame time shows 6/3 at 1501 when it's actually 6/4 at 1501.
I tried using offset but it won't go any higher than 720 minutes and I'm off over 1000 minutes.
This appears to have something to do with VMware. When I copy the Ubuntu data to a mapped drive on xp, and use that as the source for HFM the times are correct.

Thanks again

[harlam357]

Are you using the 'Client is on Virtual Machine' option?

FAH reports frame times in UTC based on the system time and UTC offset.

So, the 'Client is on Virtual Machine' option tells HFM that the client knows the local time, but thinks the UTC offset is 0 (instead of, in your case, -7). I'd have to guess that's what's going on. I know, the option isn't very clear... I mimicked what FahMon calls the same option. However, it's not as forgiving as FahMon about using it when it doesn't need to be used. There is no option in HFM to deal with asynchronous clocks, which is what FahMon appears to think is going on when using the option on a client that does not need the option specified. I think the allowing asynchronous clocks lessens the ability to accurately detect a misbehaving client.

Anyway... log files are log files... if you moved them to your local machine and tried to point HFM at those and they read correctly (i.e. showed Green). Then yes, something is up with the way that client is configured. It's got to be the VM option. If I were to subtract 7 hours from the Time Of Last Frame (Date) in the output above. It would then give you a time stamp of 08:01. Forget about the date, it is inferred from the local machine's date. It is being set to 6/3 because the frame time stamp is being read as 15:01... and local time is 08:16. If a frame time stamp is that much greater than the Retrieval Time (Date) time stamp... the frame is deemed to be from the day prior.

I can easily create the same issue... my UTC offset is -5 (CDT)

Code: Select all

[6/4/2009-7:03:36 PM] + HFM.Instances.ClientInstance.DetermineStatus (Ubuntu VM SMP - DS3P - 1)
[6/4/2009-7:03:36 PM] +  - Retrieval Time (Date) ------- : 6/4/2009 7:03:36 PM
[6/4/2009-7:03:36 PM] +  - Time Of Last Frame (TimeSpan) : 23:56:01
[6/4/2009-7:03:36 PM] +  - Offset (Minutes) ------------ : 0
[6/4/2009-7:03:36 PM] +  - Time Of Last Frame (Adjusted) : 23:56:01
[6/4/2009-7:03:36 PM] +  - Prior Day Adjustment -------- : True
[6/4/2009-7:03:36 PM] +  - Time Of Last Frame (Date) --- : 6/3/2009 11:56:01 PM
[6/4/2009-7:03:36 PM] +  - Terminal Time (Date) -------- : 6/4/2009 6:44:20 PM
[6/4/2009-7:03:36 PM] + HFM.Instances.ClientInstance.ProcessExisting (Ubuntu VM SMP - DS3P - 1) Execution Time: 141 ms
[6/4/2009-7:03:36 PM] - HFM.Instances.ClientInstance.Retrieve (Ubuntu VM SMP - DS3P - 1) Client Status: Hung

[weedacres]

Are you using the 'Client is on Virtual Machine' option?

Yes I reviewed the settings on all 4 VM's and all had the Virtual Machine Option checked.
So,, just for something to do I unchecked them and it is now working fine. HFM is getting it's data directly from each VM.

[harlam357]

Excellent! Again, I know the option isn't very clear. It really shouldn't be specified as 'Client is on Virtual Machine'. That's why it is follow with '(and reports UTC as local time)'. Because that is actually what checking that box accounts for. If a VM can be set to the correct local time with the correct UTC offset, then this option does not need checked.

Glad it's working for you now.

[weedacres]

Excellent! Again, I know the option isn't very clear. It really shouldn't be specified as 'Client is on Virtual Machine'. That's why it is follow with '(and reports UTC as local time)'. Because that is actually what checking that box accounts for. If a VM can be set to the correct local time with the correct UTC offset, then this option does not need checked.

Glad it's working for you now.

Well I was certainly confused, but all is now well. Thanks for your help.

I do have a couple of questions.
How are the Complete and Failed numbers arrived at? One of my gpus has 2440 completed wu's and HFM shows 0. I have 8 clients running and all are low numbers.
I know that Nvidia gpu folding and .net do not get along well, and I've seen where you've made some changes back in Build 13 (I think,) but in order to make any program changes or menu selections I have to kill the gpu client(s) to get any usable response. Are we stuck with that or do you have anything else up your sleeve to fix response issues.
Again great program for lots of reasons.

[ChasR]

Complete and failed are parsed from the fahlog.txt file.

[harlam357]

How are the Complete and Failed numbers arrived at? One of my gpus has 2440 completed wu's and HFM shows 0. I have 8 clients running and all are low numbers.
I know that Nvidia gpu folding and .net do not get along well, and I've seen where you've made some changes back in Build 13 (I think,) but in order to make any program changes or menu selections I have to kill the gpu client(s) to get any usable response. Are we stuck with that or do you have anything else up your sleeve to fix response issues.
Again great program for lots of reasons.

Those counts are arrived at by looking for FINISHED_UNIT or EARLY_UNIT_END or UNSTABLE_MACHINE strings in the FAHlog.txt file since the most recent client start. So yes, every time you restart your client those numbers will return to 0 the next time the client data is refreshed in HFM.

Humm, I don't really experience the lag issue anymore on my main box where I run HFM. Yes, it does have to do with nVidia drivers and how the Framework handles drawing the controls (i.e. the grid, text boxes, labels, etc), you are correct.

To be honest, I really thought this had become a non-issue... and I think that is the case for most users. But maybe lower end GPUs still have difficulty. I'm running a GTX260-216 on my main box.

What OS are you using and what is your primary GPU?

[weedacres]

Those counts are arrived at by looking for FINISHED_UNIT or EARLY_UNIT_END or UNSTABLE_MACHINE strings in the FAHlog.txt file since the most recent client start. So yes, every time you restart your client those numbers will return to 0 the next time the client data is refreshed in HFM.

Got it,, thanks.

Humm, I don't really experience the lag issue anymore on my main box where I run HFM. Yes, it does have to do with nVidia drivers and how the Framework handles drawing the controls (i.e. the grid, text boxes, labels, etc), you are correct.

To be honest, I really thought this had become a non-issue... and I think that is the case for most users. But maybe lower end GPUs still have difficulty. I'm running a GTX260-216 on my main box.

What OS are you using and what is your primary GPU?

I'm running XP32 with 2 8800gt's on a Q6600 running 3.1ghz. I have 2 linux smp's running under vmware and 2 gpu2 clients running.

If I go to HFM clients and add client it takes over 5 minutes to open the Add/Modify screen to open all the way (I gave up after 5 minutes so don't know the precise time). During that time the 2d graphics system slows way down. By that I mean while typing this note I'm experiencing about a 1/2second delay between keystroke and the letter appearing on Firefox. When I close the gpu2 client for that monitor (the one HFM is on) it instantly returns to normal and the Add/Modify window is fully open for input and the typing delay (which is on the other monitor) disappears.

[Rum@NoV]

Great work so far, i like the information about the number of failed and completed units. The UI is also very decent and i love the html output!

Is there maybe a possibility to import the benchmarks from FahMon to HFM.NET?

[harlam357]

Hi Rum@NoV... I doubt I'll be able to import benchmark data from FahMon. Although the system is similar that data is much different. I'll have to look at the FahMon code to see how the benchmark data file is built. If I can replicate the file layout in C# I'll give it a shot. However, I'm not making any promises.

If you just leave HFM running it will catalog up to 300 frames per client per WU project. GPUs obviously run up to this maximum pretty quickly... the only number that persists beyond that rolling count of 300 frame times is the minimum frame time.

Thanks for your interest! I'm glad you're enjoying HFM.

P.S. Keep a look out for the next release (shooting for Saturday)... I've made some nice upgrades to the Summary web page as well as a few other minor tweaks to the main application.

[harlam357]

New Version: 0.2.1 - Build 23 - Beta - Download

- md5: 21e8e4690bf1960f21737ba9858b8772

- sha1: c2c52c12e257052906ada6a026dc61671561a8dd

Change Log

• Enhancement: Major upgrade to the HTML Summary page. Now outputs a table with the same columns as the main HFM UI including the proper status color.
• Enhancement: Add Mobile Overview (mobile.html) and Mobile Summary (mobilesummary.html) Web Generation.
• Enhancement: Allow website to be uploaded via FTP when a FTP style URL is input as the Target Folder (i.e. ftp://username:password@ftp.com/FolderName/).
• Enhancement: Added ToolTip for the Status column that shows the text representation of the current Status.
• Enhancement: Added check of Username and Team vs. the configured Username and Team (will show Orange with ToolTip if incorrect).
• Enhancement: Added Preferences setting to control how many decimals are shown in PPD calculations (0 to 5 decimal places).
• Change: Add the user input processor MHz as a column.
• Change: Remove the Team column and Combine it with the Username column (will require you to reset your column widths).
• Change: Reference the correct casing of the CSS files, and add code to make sure previous versions of HFM will still find the right CSS selection... however, to "fix" the entry in the Preferences the user will need to accept the Preferences Form after upgrading. This change for the casing is necessary to support running on Mono in a Linux environment.
• Fix: Preferences Dialog - add code to allow only digits to be input in either Minutes field, the EOC ID, Team Number, or Proxy Port field.
• Fix: Bug that caused the main grid Sort Order to be lost after minimizing the main UI window.
• Fix: Web Generation code now wrapped in try/catch, if the web generation would fail the timers would not be reset and thus HFM would stop refreshing the client data.
• Suggested Setting (Linux): Try "xdg-open" for logfile viewer and file explorer.

Upgrading HFM to the latest release

• When upgrading HFM, please take the new Zip file and extract it to the same location as your current install and overwrite the older files.
• The user settings are tied to the location of the HFM.exe file. So if you move or otherwise rename any directory in the path to the HFM.exe file, you will lose your custom settings. Please choose your install location wisely, I recommend a very generic folder name like "\Program Files\HFM". Settings will also not follow the application from machine to machine. It is possible to transfer them and I will post instructions on how to do so at a later time.

Donate to the HFM.NET Project with PayPal

[harlam357]

New Version: 0.2.2 - Build 28 - Beta - Download

- md5: 11c4be2423fb4119fd1ee43c538c2640

- sha1: d1cb4e94f3ecba03d85c7d5f9fbfc6526ce15839

Change Log

• Enhancement: Add User Stats Data Display from EOC XML along with option to enable or disable the display.
• Enhancement: HTML Summary - User Name cells as Orange background when User Name does not match the User Name configured in the Preferences.
• Enhancement: Added five new style sheets for more Website color options.
• Enhancement: Conversation time of HTML Upload via FTP is now being logged (Info Level).
• Change: Each HTML page now has a link to the HFM.NET Google Code page. Using alternating colors on Overview page.
• Change: Small tweak to the LogParser.cs when CheckForProjectID() writes the warning message "Failed to parse the Project (R/C/G) values from '{0}'" generated by SetProjectID(). There was no indication of which client log failed the parse and what value generated the warning.
• Change: Protein Information on the Benchmarks Form now in text boxes (just aesthetics). Fix the spelling of "Preferred" on the Benchmarks Form.
• Change: All Forms are now being set with DoubleBuffered = true in an attempt to further alleviate UI lag issues.
• Fix: Grid columns not appearing in default order on some users systems. Manually create the columns and bind. Allows column headers to be labeled in a more visually friendly manner (i.e. they can have spaces now).

Upgrading HFM to the latest release

• When upgrading HFM, please take the new Zip file and extract it to the same location as your current install and overwrite the older files.
• The user settings are tied to the location of the HFM.exe file. So if you move or otherwise rename any directory in the path to the HFM.exe file, you will lose your custom settings. Please choose your install location wisely, I recommend a very generic folder name like "\Program Files\HFM". Settings will also not follow the application from machine to machine. It is possible to transfer them and I will post instructions on how to do so at a later time.

Donate to the HFM.NET Project with PayPal


HFM.NET - 4th of July Edition. Hope everyone has a safe and happy 4th! The main feature addition here is the EOC Stats display in the HFM UI. This should update on schedule with EOC Stats (hopefully this will work fine for those outside the US- timezone issue). There is an option in the Web menu to Force a Refresh of this data if you find it's not updating. Couple other small updates to the Web Generation code and pages. Lastly, a fix for the default column order on some machines. Users who have not experienced this bug will not be effected. The side effect of this is now I can label the column headers with a more friendly name as they are no longer being auto generated by the class definition.

Thank you again to the following people who have donated to the HFM.NET Project (no particular order)

• David Fonda
• Michael Obermeier
• moltenrain
• David Dory
• George Stephen
• Zachary Glazer
• Jason Farque
• Michael Metcalf
• Charles Richards
• Timothy Hester