Reminder - old proteins will be accepted for 24 hrs.

[Brian the Fist]

Just a reminder that we will accept old structures for 24 hrs now (or 48 hrs at half-credit) so there is no need to auto-update or clear your buffers immediately. Hopefully this will help reduce the load on auto-update servers and on the main server as well.

[baja27]

why when i make upload.bat and clic yes or no for the update , the update delete all my old result
and suposed have 24hr again

[Michael H.W. Weber]

The new client/server architecture does definetly NOT allow for upload of the old structures! I had test-generated slightly more than 1000 structures of the old protein. After protein update, I tried to upload my old structures. The client was updated, the old structures deleted and then processing of the new protein commenced. I checked the stats-page exactly before and after upload - no crediting for the mentioned ca. 1000 structures. Very bad. :shocked:

Michael.

[baja27]

i have more chance ,i have copy all file before send.
again in stock

[wirthi]

Originally posted by Michael H.W. Weber
The new client/server architecture does definetly NOT allow for upload of the old structures! I had test-generated slightly more than 1000 structures of the old protein. After protein update, I tried to upload my old structures. The client was updated, the old structures deleted and then processing of the new protein commenced. I checked the stats-page exactly before and after upload - no crediting for the mentioned ca. 1000 structures. Very bad. :shocked:

Hi,

I guess you can only upload the OLD results with the OLD client. So, when the update is available and you are online, don't download the update but upload the (old) results. Then download the new version and continue.

If the client deletes the buffer when you tell it not to download the update, then that's a bug (I guess).

Am I right, Howard?

Greets,
Wirthi

[Michael H.W. Weber]

I have expected this problem and copied the whole folder elsewhere. I will now try to upload manually and then report back here.

Michael.

P.S.: At least the announced speed improvement is indeed working fine!

[Digital Parasite]

I'm sure Howard will respond to this but until he does I will let you know what I have observed.

In order to upload old protein results you need to download and install the new DF client first, THEN run it to upload your old buffered structures. If you do this, your old structs will be uploaded and credited on the stats server. I can confirm this works as I have done it.

Before, when a protein changeover happened any old proteins were invalid so the client would delete the old proteins before doing an auto-update. This functionality was not taken out of this release of the client so it still behaves that way.

Anyone who has been running offline or has buffered structures should make a copy of their filelist.txt and bz2 buffer files before doing anything else. Once the client has upgraded to the new version either manually or auto, you can copy those files back and do an upload and you will get credited.

Jeff.

[jlandgr]

If the client deletes the buffer when you tell it not to download the update, then that's a bug (I guess).

For me, when I start DF back up (when exiting it before with Q the results could not be sent due to server load), the buffered files are deleted just after checking for a new version and before asking if the new version should be downloaded. If I tell the app not to download the update, it exits (this is by design, I believe).
Is there a way to skip the check for the new version and just upload the results with the old version? I will look through the docs and the posts here to try and find such a way...
Jerome
edit: thanks Jeff for explaining how to upload the old files.

[Michael H.W. Weber]

Checked it and it did not work. The client ALWAYS looks for an update first. If available that one will be used and thereby all old work gets lost.

To my sadness, your hint of first copying the old work to another folder, then updating the client, and finally copying back the old work to upload comes a little late now. My backup was just also deleted because I was too dumb to copy it again. :sleepy:

It would surely have been clever if Howard had mentioned this procedure somewhere IN TIME. Anyway, my work is gone now - wasn't much stuff, however.

Michael.

[KWSN_robegeor]

Just lost 500,000 nonetted results. When I launched the old client (and yes, with the -u t switch!!!!!) it...

1) deleted the old results

2) started to update the client

3) pissed me off

In that order



Does anyone have a process that works????????????? Please share if you do....

[Digital Parasite]

Originally posted by Michael H.W. Weber
It would surely have been clever if Howard had mentioned this procedure somewhere IN TIME.

I think Howard forgot to remove the "delete old proteins on autoupdate" code from this release so he expected it to work properly and that is why did didn't post any instructions on how to upload old proteins.

Jeff.

[KWSN_robegeor]

Great...I just lost around 3million results then...

[jlandgr]

Originally posted by KWSN_robegeor

Does anyone have a process that works????????????? Please share if you do....

Jeff's process worked fine for me: make a copy of all the *.bz2 (except rotlib.bin.bz2), *.log.bz2, filelist.txt, let the client auto-update and crunch a few structures, exit with Q, copy the work files you saved back to the folder, restart the client.
Works for me, as I said. I really am sorry you lost so much work
Jerome

[baja27]

ho yes --- that work good




1.150.000 total saved

[wirthi]

Hi,

copied from the news section of the homepage:

"Old protein will now be accepted for up to 48 hrs. after each update. However, be aware that the client/screensaver is likely to delete old files when a new version is detected (like it always has). So you may need to back up your buffered files, then let it auto-update, then put them back to prevent them from getting erased. Or just manually install the update and don't let it auto-update. We will release a fix to avoid this shortly for your convenience with future updates."

[KWSN_robegeor]

Originally posted by jlandgr

Jeff's process worked fine for me: make a copy of all the *.bz2 (except rotlib.bin.bz2), *.log.bz2, filelist.txt, let the client auto-update and crunch a few structures, exit with Q, copy the work files you saved back to the folder, restart the client.
Works for me, as I said. I really am sorry you lost so much work
Jerome

Jerome,

I will try that with my remaining CPUs.

Thank you and

rg

[jlandgr]

Originally posted by KWSN_robegeor

Jerome,
I will try that with my remaining CPUs.

All the credit goes to Jeff
Good luck,
Jerome

[Digital Parasite]

You're welcome guys.

I have been burned by this feature before where it deletes your files before it asks you if you want to auto-upgrade y/n so ever since then I have been backing up my buffer files and it looks like it came in handy today. That let me play around with stuff to figure out how to do it successfully.

Jeff.

[KWSN_robegeor]

Thank you Jeff!

It worked great using your process.

[Brian the Fist]

The deletion feature was originally intended to benefit you so it wouldn't keep crusty old protein files lying around and then waste time trying to upload them.
Unfortunately I forgot all about that until just today, so if you follow the directions on the News page it should still work OK.
I will add a little extra checking now so that it will delete files only if it receives a 908 structure error (meaning the files are truly expired). This was what was intended. Check the news page for when this fix is available, in a day or two.
Sorry if this inconvenienced anyone.

[MAD-ness]

Lost upwards of 500,000 structures, but from what I have heard the new feature to load the full protein into RAM works great, so we will consider it a wash, alright?

[Pascal]

I lost 30,000 structures
But I'm still a bit happy, because today's my 26th birthday, I lost the structures yesterday

[Michael H.W. Weber]

Happy birthday, Pascal.

Michael.