[PCCstudent]

Maybe not sticky material but perhaps a informative thread about proper way to halt folding when you need to make a change of somekind. I would like to know under what circumstances you are guaranteed to lose what ever work you have going on and what is the "nice" way to stop a client.

I also wonder what Stanford sees when one of use is repeatedly starting and stopping a client and does it cause trouble on their end. I want to stop my 6.29 folder but all is working so well I just don't want to mess with things. I found you can 6.29 fold quiet well on just 2gb RAM so I want to pull 4 out and add it to my i7 so I can get angry trying to setup -bigadv

Things like automatic updates and a boot procedure that wont complete without something like a password get me uptight. Last night I pulled to hard on my monitor cable and dislodged my gpu and had to do a restart as I just could not get video back otherwise. I did not lose any work completed but I was worried.

 

[CryWin]

I just close it when I need to do something, and the work is resumed when I re-open the client.

 

[MadCatMk2]

Ctrl+C it. It's not going to matter to stanford. Communication only occurs for sending and receiving workunits.

 

[RonindeBeatrice]

Just close the window normally. If you're quick you can read the shutdown message before the window disappears, signaling the ok.

Under linux I just press ctrl+c and wait for several stopped messages to appear, then I shut down the VM.

Also, stanford doesn't receive your log files, so they've no way of knowing at what times you shut the WU down.

 

[k4m1k4z3]

Ctrl-C it and get that i7 going

I do it all the time to mine when I need to do a reboot. It never fails to start where it left off.

 

[PCCstudent]

OK , I was worried that during a inital configuration and WU download that did not go right (like when I kept getting Falcor_82 work) and repeatedly modified my arguments that it was taxing the server sending me work. What I conclude is that shutdowns are nothing to fear and they most of the time don't cause you to lose work. It is just when you must delete things like Fal 11 or 14 to get past a 24 hr hold that you most certainly lose out, do I have it correct?

 

[franz]

Quote:

Originally Posted by PCCstudent OK , I was worried that during a inital configuration and WU download that did not go right (like when I kept getting Falcor_82 work) and repeatedly modified my arguments that it was taxing the server sending me work. What I conclude is that shutdowns are nothing to fear and they most of the time don't cause you to lose work. It is just when you must delete things like Fal 11 or 14 to get past a 24 hr hold that you most certainly lose out, do I have it correct?

Like the others said CTRL+C is perfectly safe for SMP and GPU clients.

When you go in to start deleting things stay away from the work, queue, and unitinfo files.

The easiest way to get rid of the EUE limit 24 hour wait, is to delete your FAHlog file. That way you wont risk losing any unsent WUs.

 

[[CyGnus]]

hummm i do not know why but sometimes i loose work in SMP after a reboot.... not always but in say 10 reboots maybe 3 times a loose work and i do not change the OC just reboot...

 

[zodac]

See, I think the clients have gotten a little more stable in that respect. I remember when I started GPU folding; any change to the clocks would result in the client crashing and getting a new WU. Now I can change 3/4 shader straps and have the drivers crash, but the client still runs (card downclocks, but still).

As for what Stanford see, I think nothing. The only data we upload is that of the WU when we upload the results. Aside from the actual research, they see when we downloaded and when we completed the WU.

Automatic Updates can be disabled (handy for the Foldathon) and you can have Windows 7 log you in automatically (even with a password) so even in the case of an update/reboot you'll continue to fold I'm sure you know that, but I'll still mention it).

 

[Ruckol1]

Quote:

Originally Posted by [CyGnus] hummm i do not know why but sometimes i loose work in SMP after a reboot.... not always but in say 10 reboots maybe 3 times a loose work and i do not change the OC just reboot...

I have this problem too

 

[PCCstudent]

With my Falcor_82 debacle was it not seen that I was requesting a WU about 10 different times? They did not send me a "nasty note" so I guess it did not matter so much

 

[zodac]

I believe that WU is kept in the server and once the deadline is passed it's sent out again to someone else. I suppose they've got enough work to do with the reults they actually get back, so automate the servers to continously send out the WUs until some results come back.