Hang on Start still a problem....

[Aardvark]

Downloaded the Client6.20 from Stanford shortly after it became available.
Prior to Install, I did an uninstall of the Client6.10beta2 as suggested yesterday by Tohya in the "Mac OSX Beta" Forum:

To completely uninstall, you want to delete these 4 things...

1. /Applications/Folding@home.app/
2. /Library/LaunchDaemons/Folding@home.plist
3. /Library/PreferencePanes/Folding@home.prefPane/
4. ~/Library/Folding@home/

I did the Client6.20 Install and "initialized" things in the System Preferences Panel. The Client took off like a charm, gathering the vital parts from Stanford. Everything acted like it was a GO so I left it to grab some shuteye. I returned about 5 hours later to find it had hung at the:
mdrunner cpfilename:
output line.

I did a "manual" start/stop of the Client in the preferences panel and the Client started grinding full bore immediately.

This is a repeat of a problem that I was frequently experiencing under Client6.10beta2. I am still not aware whether this is a hardware problem, a System software problem, or a Client-Core problem ( or something else )

I am experiencing the above on a 1.83 Ghz Mac Mini running under OSX 10.5.4.

I would appreciate any work-arounds for this problem. I had hoped that it would be cleared in 6.20 but apparently that is not the case for me.

[Aardvark]

My prior posting in this thread has been up for 24 hours and there has not been so much as a whisper in response. I had hoped that my posting would at least draw some comment, if not specific directions, as to how to avoid the lost cpu time when the client stalls and does not report that fact.

Having said that, let me attempt to restate what I am looking for. Is there some way that I can FORCE the OSX Client6.20 software to be more “verbose” such that I will be informed through FAHlog.txt (or other message system) that the WU has stalled and is in need of a “nudge”? Next needed piece of information? What are considered acceptable “nudges” that will not affect the integrity of the WU data but will get the WU processing back into action?

Is it expected that OSX Folders will nurse WU startups through the wee hours of the morning, starting and restarting the Client manually with the System Preferences Panel in the hope that there is Life in this WU, and by golly “I will find it”? I think not. This Project must have Software Engineers and Programmers capable of more elegant solutions. Let them “Do it for THE CARDINAL” so that OSX Client6.20 can become a source of increased Folding productivity.

Just my thoughts as I anticipate a startup on a new WU later today. I hope someone might have some experience that can be shared.

[Tohya]

Use -verbosity 9 to get a more detailed log.

[rickduff]

Just some random thoughts.

Check with the activity monitor to confirm nothing else is using CPU running at a higher priority - I have even found the Finder to be the culprit at times (just force quit it and it will restart - the Finder anyway).

Is your machine a single core? I know some of the Mac Minis are. I am not sure what happens when you try to run it on a single core machine.

As mentioned, post a new log from startup and -verbosity 9 so we can see if there are any clues there.

Rick..

[Aardvark]

Rickduff,
Thanks for the thoughts.
When the problem has occurred activity monitor indicates "low" CPU usage. I try to keep other applications in a shut down mode just on the outside chance they might "offend" the F@H Client-Core dance.

The hardware where this problem is occurring is a 1.83 GHz Dual Core Intel Macintosh Mini.

The WU that I am currently running will finish later today. It started folding on its own after a couple of hours "thinking". Will get another startup later today. I have reset the verbosity.

Will just have to wait for more results.

[Aardvark]

I have continued to pursue resolution of the “Hold on Startup” problem I am experiencing. As suggested by Tohya I increased the verbosity setting in the OSX Client6.20 Preference Panel hoping to get the Client to talk more.

Since last posting in this thread, I have monitored the StartUp of 2 WUs. Both are from Project 2605. Both exhibited the same “Hold on Startup” problem. The WUs assemble themselves cleanly and the process proceeds rapidly to outputting to FAHlog.txt the following:
Enter M.D.
mdrunner cpfilename:
rejecting checkpoint

At this point everything seems to go on Hold. No more output to FAHlog.txt. After sitting “at idle” for approx. an hour and a half I disabled and restarted the Client in the Preferences Panel. The process again rapidly proceeds to the same output to FAHlog.txt as entered above and then went on Hold.

After approx. two hours into this process I have decided to do a complete Computer Restart. I shut down the Client and rebooted the Mac Mini . After restarting the Client in the Preference Panel it immediately reported that it was starting to fold in FAHlog.txt. From this point on, folding proceeds and finishes completely normally.

I have closely observed this behavior twice in succession. All I can say is
“Houston, we have a problem”.

Was this Client written to call for a complete machine reboot before it will proceed to folding on a new WU? I haven’t seen that written anywhere, so I think not.

What is the nominal time that is required, on this machine, before folding should begin? As I presently see it, that is the point at which a reboot is required so that folding can proceed. The machine is capable of completing a frame in 23 minutes, once it gets to it. It just doesn’t make sense to waste a couple hours at the start of a WU hoping it might decide to fold.

I know that there is a repeatable problem. I don’t know whether it is Hardware, System Software , or Client Software related. Some guidance from the gurus would be appreciated.

To refresh memories:
Hardware---Mac Mini 1.83 GHz Intel Core 2 Duo with 1 GB Ram
System--OSX 10.5.4
Client-- OSXClient6.20

[calxalot]

Did you try adding your hostname to the localhost line of /etc/hosts ?

[Aardvark]

calxalot,
I have found some discussion about doing that in a prior thread on the Forum. EXACTLY what is supposed to be used as my hostname???
I am a long time Mac user but just not familiar with some of the terminology. I suppose that whatever is entered here must be EXACT.

Thanks for the interest.

[calxalot]

Open a terminal, and type "hostname" without the quotes.
You will get something like Aardvarks-computer.local, or whatever you put in System Prefs > Sharing.
You will need to edit /etc/hosts as root.
If you are not using an admin account, you must switch to one, or use

Code: Select all

su - admin

(assuming "admin" is your admin account short name)
If you are an admin user, you can type

Code: Select all

sudo open -a TextEdit /etc/hosts

change

Code: Select all

127.0.0.1   localhost

to

Code: Select all

127.0.0.1   localhost Aardvarks-computer Aardvarks-computer.local

Quit TextEdit immediately when done. You don't want to edit anything else as root.

[Aardvark]

Thank you for the info, calxalot. I have just started a new WU with the same problem. Had to reboot the Computer after three software induced client restarts to bring life to the WU. I have to believe that what you are suggesting will solve the problem.

Should I edit the /etc/hosts file between WUs or can I do it at any time assuming I disable the Client, edit the file, and then restart the Client during the current WU?

[Aardvark]

calxalot,
I did a quick trip to Terrminal and entered "hostname" (no quotes),

The response was:

136.220.61.64.modem.dynamic.cptelecom.net

I am on a 56k dialup access. When I have raised that as a possible cause for other problems, the response has been "It won't be a problem".

I am assuming this hostname would not be suitable for the /etc/hosts file modification. Am I right? If I am, is there any other workaround possible? I suppose this might be why we have proxies but I have never gotten involved in one of those.

[calxalot]

Use what you see in Sys Prefs > Sharing.
If that doesn't fix it, try adding the long name you got above.
You can edit /etc/hosts whenever you like, but the client will probably need to be restarted.

BTW, the SMP client will crash and delete the current WU if the network changes.
It's one of the known issues for the smp client.
I don't know if a dialup link changing would also cause this.

[uncle_fungus]

It looks to me like your machine name is being determined by the dhcp lease from your ISP. Since it is highly likely this will change every time your machine reconnects to the internet this will cause problems as your hostname will be different.

Is there a way in OS X to specify a fixed hostname (I couldn't find one) rather than assigning it from dhcp? If so, it should be done, and then that fixed hostname should be entered into your hostfile.

[Aardvark]

I have tried to modify the /etc/hosts file usiing the hostname available in SysPref > Sharing , as suggested by calxalot. I had the file edited and was not allowed to save it due to "Lack of Permission". As I understand it, just because I am Admin according to OSX System I am not necessarily Admin(root) according to UNIX(or variant thereof). I am the sole user of this machine and I set the System Software up after purchase.

I think I have seen something, somewhere about establishing yourself temporarily as "root" so these administrative tasks can be done.

Uncle_Fungus also raises a troubling point which I don't have the answer to. I just assume that if the /etc/hosts file can be modified with the proper local information that this will satisfy the Client-core startup needs. I expect that the client will go to this file rather than something provided externally by the ISP. Does anyone agree?

[jackrabbit]

Rather than trying to get TextEdit to work as root, use this command:

Code: Select all

sudo nano /etc/hosts

It opens an editor inside the Terminal with a usable UI. Arrows work as usual. There are some lines at the bottom which list the commands. ^X means press the control key and the x key at the same time.