ghclient many pcs [Archive] - The Genome Collective

viperdog

25th August 2001, 18:44

just some notes...

the ghclient only saves it's work at the end of each of the 30 sequences it runs, so try not to shutdown until it gets to 100%.

If you do shutdown, on restart it will re-run the filter section and then continue where it left off e.g. shutdown at 20% seq 6, restart, filter seq, begins at 0% seq 6.

You will notice it tries to download a work unit to cache if
a. it doesn't already have one
b. only on a restart with an unfinished work unit..

if you are not on the net then restart with -nonet will start process right away, otherwise it will time out the net connection wait 2 min try again ( 3 times I think) and then start to work on the unfinished unit.

Same thing will happen if it gets to the end of a work unit and is not in -nonet, it will try to connect..fail in about 6 min and re-run the prior unit (with a modified variable to produce different results). Next time you upload all the wu you have processed will uplaod at once.

Vermine

25th August 2001, 22:49

To know exactly what file to copy on the nonet computers and to the computer on the net... Here is a FAQ that i've found on [H]ardOCP team forum:

Q. Which files should I copy from a non-connected machine to upload my results?

A1. To be safe, copy everything from the no-net machine to a FRESH G@H directory on the connected machine - except filter.exe and ghclient.exe. You don't need these, as they are the executable files and are not altered from instance to instance. Make sure you don't copy over files with the same name from different genes!!!

A2. Due to the confusion surrounding the proper way to hoard files for the impending dump/Ex-Lax orgy, I conducted a test last night of a mini-dump to see if an idea I had would work. IN SHORT, IT DID WORK, and very well I might add . The steps I took are outlined below:

1) I had been running the majority of my clients -nonet and harvesting the results every three or four days. To harvest the output for Stanford, I would backup, by cutting and pasting, all csum.###, output.chi.###, and str.### files in a separate, central storage directory. Those are the only files in that storage directory. Occasionally, while cutting and pasting results, I would get a warning that the .### extension on the files being pasted matched that of files already existing in the storage directory - WARNING DO NOT PROCEED TO OVERWRITE THE EXISTING FILES WITH THE ONES BEING PASTED INTO THE STORAGE DIRECTORY. If you do you will destroy some of you output that you have worked so hard to create. What you should do is cancel the paste function and instead, paste them into a separate subdirectory within your storage directory which will be used for files with duplicate .### extensions.

2) Continue to harvest your output as described above until you are ready to ingest the Ex-Lax.

3) When it is time to dump, create a separate directory for G@H on your HDD, I suggest C:\Dump Station. Within that directory you want to copy the following files from the directory you have G@H installed into:
http://www.orangedelta.com/images/gah.jpg

NOTE: There is no ID.DAT file. If you have one, delete it. You will also notice that these are just the basic files needed to run G@H. Make sure your ghclient.cfg file is correct, ie it has your user name and our team number: 2136075668.

4) To setup for your dump, copy over all of the csum.###, output.chi.###, and str.### files you have in you main storage directory into your C:\Dump Station directory. DO NOT copy the duplicate csum.###, output.chi.###, and str.### files in your storage subdirectory at this time unless you want to overwrite some of your output. Alos, make sure you save a copy of these files in case Stanford gets overloaded or there is a server outage or electrical outage (afterall, we are talking about Califorinia!).

5) To actually make the dump is simple: simply run the client executable [ghclient.exe] without any flags [I double-clicked directly on the executable itself b/c command line DID NOT work]. It will connect the the Stanford servers and upload all of your output files, ie the csum.###, output.chi.###, and str.### files you have stored, and then begin to look for new work to download. Many of you are prolly screaming that you have to have an ID.DAT file to get credit. WRONG. That simply IDs a particular machine. Once you start your upload, the client will generate a new ID.DAT file for you so it is no big deal. It may not even matter at all, but I did not test that.

6) Once the inital dump is done, then copy over all of the duplicate csum.###, output.chi.###, and str.### files you had stored in the storage subdirectory into the C:\Dump Station directory and repeat from step 5.

7) Viola! You are done!