Mojo
14 August 2004 09:54
1
NEW CLIENT 3.22
Please read the note before to download and install the new client
If you are experiencing the problem that your client(s) runs applications with the wrong files and you want to help us finding a solution, download the new client 3.22 for Windows here. If you are confident with the MSVC program database, you can download these additional files. If the new client continues to mix up the applications, send us: sched_reply.xml, sched_request.xml, client_state.xml., stderr and stdout as well as the log files produced with the MSVC program database.
The new client 3.22 is currently only for test purposes. Please do not download it if you do not want to challenge your machine but keep crunching with the client 3.20. We cannot guarantee that the client 3.22 is the solution to the current problem but it will help us to fix it.
The current client 3.22 is only for Windows. A Linux/Mac version will be available soon.
This is obviously a new version with more debug to help them track down the error they are experiencing. If you haven’t the time to monitor the client or send the data back to Predictor then I’d leave it alone.
8/21/04 - Update of the day:
This is great news directly from David Anderson: "We located and fixed a bug in the feeder that would occasionally cause results to be associated with the wrong workunit. In cases where this workunit belongs to a different application than the result, this would usually lead to a confusing situation on the client, where the application would fail to produce the expected output files. In any case, the redundancy mechanism would generally detect and reject the bad result, so the only damage done is that the user doesn’t get credit. "
We have briefly tested the new feeder on our local testbed and upgraded the daemon on Predictor. You do NOT have to reset the project or upgrade the client. Just go on keeping us updated if you see any error message on your client.
Special thanks to: David who has found and fixed the bug and all the P@H users that have provided us with the client log files and patiently are crunching P@H despite some minor problems.
Basically I take that as all those who did move a rig over and provide log files have assisted in sorting the problem, with David Anderson being able ot pinpoint and resolve the issue. What was happening was some mbfold units etc etc were trying to compute using the charm set of code and vice versa.
As news post says, no need to update, another plus of the BOINC project, the update will be done for you when needed.
DT.