I upgraded to 4.59 so far with success. If I need to go down to 4.15 I will. I finally got my e-mail inviting me to join the test and am happy to be doing so.
> Is this upgrade necessary if the host is not running CPDN?
The bug that is fixed concerns possible loss of work when starting the benchmark (which happens e.g. when you upgrade the client, but also about once a week). In principle it affects all projects, and every time the benchmark is ran gives a possibility for breaking the current project. It turned out that the CPDN app currently is the one most sensitive for this bug, and because of the very long time the CPDN WUs take, it also does the most damage to this project. But sooner or later every project may be affected.
I too got the same error that Juerschi got from my Predictor unit. Did the prudent thing and re-installed 4.14 and everything is running fine again. I guess I'll wait to see what the problem is about before going back to 4.15.
In the Alpha_Dev mailing list, Rom Walton explains:
"An issue was detected where cliemateprediction.net workunits were failing
during the upgrade process.
Basically the bug fix for this release, checks to see if benchmarks need to
be run before any of the science applications are started, before BOINC
would start the applications, and then detect if benchmarks need to be run.
If people had the flag enabled to remove the process from memory that means
the application could have received the quit request before it had even
completed reading its state files.
We hope this will clear up the CPDN workunit failures."
---
This bug fix decription wouldn't explain the sudden problems with Predictor.
PS I enabled the "keep in memory" option some time ago.
> PS I enabled the "keep in memory" option some time ago.
I think anyone running CPDN uses this option. (at least those also doing work for other projects) Otherwise, they wouldn't get anywhere with the WU because the checkpoints are somewhere around 8 hours apart.
I was already running Boinc 4.59 ( attached to seti+Predictor) on a PIII 1GHz WinMe.
Last night I attached to Einstein. It downloaded 8 WU's, 1 has finished, was uploaded and reported... second WU started crunching.
so far so good using einstein 4.71 and BOINC 4.15. BOINC 4.59 was not able to do einstein on my computer. (1.8 GHz, 496 RAM, intel celeron, windows XP, SP1) The CPU total time or time to completion disappear whenever einstein pauses and reappear when it is computing. Predictor and SETI keep the time up all the time.
Ed Griffith
I upgraded to 4.59 so far
)
I upgraded to 4.59 so far with success. If I need to go down to 4.15 I will. I finally got my e-mail inviting me to join the test and am happy to be doing so.
> All my clients (well, just
)
> All my clients (well, just two) are already upgraded to 4.14 with no CPDN
> loss of data. Is it still a good idea to upgrade 4.14 to 4.15?
I have also been running 4.14 with no harm done to CPDN.
Edit: Have also just upgraded to 4.15 and CPDN is still banging along.
> Is this upgrade necessary
)
> Is this upgrade necessary if the host is not running CPDN?
The bug that is fixed concerns possible loss of work when starting the benchmark (which happens e.g. when you upgrade the client, but also about once a week). In principle it affects all projects, and every time the benchmark is ran gives a possibility for breaking the current project. It turned out that the CPDN app currently is the one most sensitive for this bug, and because of the very long time the CPDN WUs take, it also does the most damage to this project. But sooner or later every project may be affected.
BM
BM
To those in charge: I too
)
To those in charge:
I too got the same error that Juerschi got from my Predictor unit. Did the prudent thing and re-installed 4.14 and everything is running fine again. I guess I'll wait to see what the problem is about before going back to 4.15.
WCK
Got all "unrecoverable
)
Got all "unrecoverable errors" on BOINC 4.59 for all work units so am switching to BOINC 4.15. I will let you know how it goes.
Ed Griffith
In the Alpha_Dev mailing
)
In the Alpha_Dev mailing list, Rom Walton explains:
"An issue was detected where cliemateprediction.net workunits were failing
during the upgrade process.
Basically the bug fix for this release, checks to see if benchmarks need to
be run before any of the science applications are started, before BOINC
would start the applications, and then detect if benchmarks need to be run.
If people had the flag enabled to remove the process from memory that means
the application could have received the quit request before it had even
completed reading its state files.
We hope this will clear up the CPDN workunit failures."
---
This bug fix decription wouldn't explain the sudden problems with Predictor.
PS I enabled the "keep in memory" option some time ago.
- Marco
Team Canada
> PS I enabled the "keep in
)
> PS I enabled the "keep in memory" option some time ago.
I think anyone running CPDN uses this option. (at least those also doing work for other projects) Otherwise, they wouldn't get anywhere with the WU because the checkpoints are somewhere around 8 hours apart.
I was already running Boinc
)
I was already running Boinc 4.59 ( attached to seti+Predictor) on a PIII 1GHz WinMe.
Last night I attached to Einstein. It downloaded 8 WU's, 1 has finished, was uploaded and reported... second WU started crunching.
No problems until now. :-)
just get the Version
)
just get the Version 4.72
Einstein@Home - 2005-01-19 19:35:50 - Started download of einstein_4.72_windows_intelx86.exe
Einstein@Home - 2005-01-19 19:35:50 - Started download of einstein_4.72_windows_intelx86.pdb
Einstein@Home - 2005-01-19 19:36:43 - Finished download of einstein_4.72_windows_intelx86.exe
Greetings from Germany NRW
Ulli
[img]http://home.teleos-web.de/ubrinkschmidt/smile/xyxwave.gif [/img]
so far so good using einstein
)
so far so good using einstein 4.71 and BOINC 4.15. BOINC 4.59 was not able to do einstein on my computer. (1.8 GHz, 496 RAM, intel celeron, windows XP, SP1) The CPU total time or time to completion disappear whenever einstein pauses and reappear when it is computing. Predictor and SETI keep the time up all the time.
Ed Griffith