New Protein Change

[IronBits]

Distributed Folding Project
Server temporarily down while we change to a new protein. Please check back in an hour.
Just got that message...
keep an eye on your clients! ;)

[magnav0x]

The new protien and client are available now, the auto update worked like a charm.

[Paratima]

Thanks for the alert, IB. Two of my boxen running Win98 DIDN'T auto-update. Had to be hand-fed. Will check on my Win2K & NT boxen in the morning.

[Dyyryath]

Bah, mine don't appear to have auto-updated either. I'm really tired of reloading clients. :(

I'm still not done reloading all my clients from the 2000 different versions we saw over last weekend. I'm not sure I'm even going to bother. Maybe I'll go back to a different project for awhile...

[magnav0x]

well, ubero could use some help, I put my client over there for a bit, but we have fallen far back into 3rd on that project, I just ran DB for a bit to get my posistion back from Steve :P

[magnav0x]

and yes when I say auto update worked fine, I mean....I manualy shut down the client while the server was down for protien change, then ran it again when the server was ready again and let it auto-download the new client.

[IronBits]

I can't find the new DF client to upgrade to!
If you follow the links and get the client from the projects download page, the dates are 3-14, which is the old version :(

I will be moving many of my computers to another project for awhile...

I would love to run Ubero, but, that's not really a project.
They are not using the DATA, so it's just wasted cycles. They are taking away computer time that could be put to good use on a real project. They have done enough testing I think and they should take it off-line until they have some real work to do. IMHO.

I personally find it hard to get excited about or feel proud of, the amount of cycles that my computers generated, knowing they were being dumped and wasted.

To each his own tho, I'm not knocking anyone.

[Dyyryath]

Refresh your browser, young Jedi! :D

They're dated Mar 24, though they've actually been up most of the 23rd.

I downloaded them around Saturday afternoon and loaded a handful of clients. They are happily crunching away and turning in work without problems.

[IronBits]

I did Master, I swear! ;)

http://www.dbestern.net/files/clip0001.png

[ulv]

IB- I did the same thing with the last update, downloaded three times, still had the old client, refreshed the browser and there..... the new client emerged on my screen...

Anyway, none of my three boxes autoupdated, one was down, two others were crunching old work. Had to download and reinstall. It doesn't do much harm for me because all my computers are at home, but this autoupdate must work soon. I still crunch G@H, very stable for me.

[IronBits]

Linkage???????
ftp://ftp.mshri.on.ca/pub/distribfold/download/distribfold-current-win9x.zip

That is the wrong version, dated 3-14, I refreshed, I shift refreshed, I shift-ctrl-refreshed, I shift-ctrl-alt refreshed.

[pointwood]

The client on my Win2k notebook updated perfectly.

This is still a very new client and there will be more new releases with bugfixes and new features. That is just the way it is.

Howard has stated from the start that the autoupdate feature was untested and might cause problems. The first time it didn't work very well. He fixed several problems that occured and I believe this latest update was/is much more smoothly. However, some people seems to still be having problems. Those that are experiencing problems, *please* report those in the bug database on the yahoo group or post a clear description here and I'll add it for you. if Howard doesn't get those bug reports, he can't fix the bugs!

[dnar]

If you create a text file called "autoupdate.cfg" in the client directory, with the content "1" (a single one) then the client wont stop and ask you if it should update the client! Great for clusters and farms.

[pointwood]

So far it sounds like it is only Win9x/Win98 machines that are having problems with the update.

Have anyone had problems with WinNT/2k/XP or the Mac or Linux/Unix versions?

[Paratima]

Have several boxen running Win2K; none of them auto-updated, although the autoupdate file was there. Most of them got multiple instances of "can't connect to anteater2...", the server doing the update. One of them actually got through and got an "update not signed" or words to that effect. In all cases, they just went on happily crunching the old protein!

Note: This is a mix of running as a service and as just another process.

[whilden@ars]

-Pointwood-

My stats #'s havent changed overnight so that means a Windows XP workstation and an NT4 server didnt auto-update. So we can squash that idea.

mike

[xj10bt]

From what I've looked at so far, no machine I have autoupdated, running either as command line (with autoupdate.cfg) or service. Restarted clients, many instances of "digital signature not found" followed by client shutdown. After serveral retries they would update. Had a handful restart and update successfully the first time. Still have orphaned work from last run.

I'll get my machines back up and running one more time, but if the next switchover isn't seamless, that's it for me.

[Angus]

It appears that none of my boxes auto-updated, and they are all text clients with the autoupdate.cfg file in place.

I came home last night to see the two Win98 boxes here happily crunching away on the OLD protein - no sign of an attempted autoupdate.

I closed the clients and restarted them. One downloaded the update OK, but choked when the install program tried to un-zip it, with an 'unable to write foldtrajlite.exe' error. Would this be because it is running at the time, and it won't replace an open file?

The other Win98 refused to get a digitally signed update after many tries. I finally downloaded the update manually, and installed it. It still tried to download, so I must have gotten the 'old' version folks were complaining about. The second manual download worked. However, that box quit crunching when the regular blank screen saver came on sat there all night doing nothing. When I moved the mouse this morning, it continued happily where it left off, about 300 structures into it's first batch. Ugly.

It also seems that none of my NT4 clients at work auto-updated since the only overnight results I got were from the one box here at home.

Not a successful operation at all
[edited by IronBits with permission from Angus]

[ColinT]

Angus:

I saw what you saw, but in addition, I see boxes simply stop crunching for no good reason.

[whilden@ars]

-Colin-

I also saw that!!

I was testing how the changeover would do on my home machine. And this machine did pull the new protein from a fresh restart, however stopped on me 3 times after folding a few WU's.

Went back to Sengent and Prime on this machine. At least I know no wasted cycles.

[dnar]

The Linux client is flawless.

[Moogie]

Originally posted by pointwood
So far it sounds like it is only Win9x/Win98 machines that are having problems with the update.

Have anyone had problems with WinNT/2k/XP or the Mac or Linux/Unix versions?

The config file didn't work for me. :( I had to download the new client and start her up again. Seems to be doing fine. I have Win2K.

[bwkaz]

Originally posted by dnar
The Linux client is flawless.

Yes, it most definitely is. I had to shut it down yesterday to run it on a VT instead of taking up an xterm, and then again to run it from inittab on boot rather than a shell (think 2K service here rather than manual startup -- if you want to do this, add "fold:2345:respawn:/path/to/foldit </dev/vc11 >/dev/vc11" to inittab, replacing vc11 with whichever virtual console is free and you want to run it on, and then "kill -HUP 1" to start it, or just reboot). When I restarted it the first time, it said "update available, want to get it?" One "yes" later, it downloads, installs, and starts crunching new proteins.

:cool:

[pointwood]

Yearh, looks like I was too fast there - when I got to work, a Win2k machine and machine was still happily running the old version. When I restarted it, it said a new version was available, but it failed to download it with the error: "Authenticy failed" :(

[tomoose]

Originally posted by bwkaz


Yes, it most definitely is. I had to shut it down yesterday to run it on a VT instead of taking up an xterm, and then again to run it from inittab on boot rather than a shell (think 2K service here rather than manual startup -- if you want to do this, add "fold:2345:respawn:/path/to/foldit </dev/vc11 >/dev/vc11" to inittab, replacing vc11 with whichever virtual console is free and you want to run it on, and then "kill -HUP 1" to start it, or just reboot). When I restarted it the first time, it said "update available, want to get it?" One "yes" later, it downloads, installs, and starts crunching new proteins.

:cool:

Could you run this over me again in a little bit more detail please. I have been trying this all today - it just seems to ignore me, no error messages, just the client doesn't run. Do I actually *have* to restart for it to take effect? The only other question about this I have is that will something in inittab with the layout you suggest above (/path/to/foldit) run withouth needing a ./ in front? (I am a linux newbie to this). Also do I *have* to specify a tty and if I do does it have to not appear anywhere else (I'm looking at the set of 6 mingetty things in the inittab file) Any help would be much apprichiated :bang:

Thanks

tom

[bwkaz]

I don't actually run it directly from inittab anymore, I run it from the boot scripts now. The main reason I don't run it from inittab is that there's no clean way to shut it down if that's how you start it. If you run it from a boot script, you can rm the foldtrajlite.lock file in the shutdown script, and wait (using /bin/sleep 2 or similar) until the foldtrajlite process is gone (check with ps -C foldtrajlite >/dev/null).

But from what I remember, that line was all you needed to run it from inittab. Open up inittab in an editor, add that line, save, exit, and do a kill -HUP 1 to get init to re-read inittab and start up foldit. You have to do all this as root.

The reason you normally have to specify the ./ in front of executables is because the shell needs to know where in the directory tree they live. The . directory is an alias for "the current directory", so it looks there. If you use the full path instead (note that you do have to replace /path/to with the actual path -- on my setup it's /home/bilbo/distribfold-icc/foldit, but on yours it will be different), you don't need a dot, because the shell (well, actually, init in this case) will know where in the directory tree to look for the executable.

I really don't know what will happen if you don't specify a TTY. Everything else in that file except the init scripts does specify the tty it wants to run on, and the init scripts seem to just run on whichever one is active. I believe that you therefore must specify a tty if you don't want the client to interfere with anything (like the shell you're trying to use on VC 1). It will help a lot if it doesn't appear anywhere else -- but it's easy enough to come up with a new one, just add one or two to the biggest number you find in inittab (which is probably 6). You add one if you don't want to run X, and two if you do (X uses its own virtual terminal device, which is selected by it scanning through which ones are used until it finds an open one -- I think it's easier to keep X on VT 7 in your case, 10 in mine, and run the client on 8 or 11).