client 4.1x, do we need aborting a single Work Unit?
2 Feb 2005 19:39:42 UTC
Topic 187430
(moderation:
Hi!
Alpha client are offering the more advanced functions like to suspend or abend
a single Work Unit.
The so called public client version (until 4.19) are "only" allowing the user
to RESET the full project :----(
If my poor understanding is right, the resetted work is lost and a *big* download
has to be restarted. Probaly the deadline must be passed until this lost work unit(s)
are declared as "lost"
This is sad.
I do know, the projects dev and admin team is doing more than their best efforts
to figure out and eleminate those candiates who let suffer our brave science-crunchers.
One of my hosts, a fellow AMD2800 (cc4.19) has fallen into the "never ending trap".
Now there is a tool, limited to Einstein Work Unit, "able" to do the single abort of a
Work unit, working with regular (-4.19) m$ Clients;-).
If there is an interest, I will put a link for download the 41 KB sized VB tool.
First listening to feedback, I would never do something "bad" to the project...
A coins has 2 sides..
Friendly!
===============================================================================
the readme:
Use at your own risk!!
=========================
Function/target: Abort a Single Work Unit for 4.16 - 4.19 M$ Clients
Step -1-
Look to the client with the "not finishing" Workunit
notice the name of the Work Unit
Step -2-
Suspend and exit the client
Step -3-
Put this tool in the same folder as the boinc client is installed
Start the tool and click "get info"
In the list box the active Einsteins tasks are displayed
Now double click the line showing the to be aborted WU
A last Message Box with the WU Name is displayed, click "OK"
watch the listed steps.
A Backup of the client_state file is done.
Step -4-
close this tool
Now start the boinc client first time
An "Unrecoverable error for result H1_..Test02_1 (Output file error: -1)" is created
Suspend and exit the client
Now start the boinc client for a 2nd time
The "not finishing" Workunit is ready for Reporting
and the next Work Unit in Queus should have been started
client 4.1x, do we need aborting a single Work Unit?
)
Make it easier, go to the slot directory, you see the big files, backup the slot where the bad WU is then delete all in the slot what you have backuped. Now start Boinc, an computation error will come and the next Wu will start to crunch ;)
You are right, this will work
)
You are right, this will work too.
Your mentioned way will generate this error:
Unrecoverable error for result H1_0184.9___Test02_1 (Couldn't restart the app for this result: -108)
Every effort to avoid to reset the full project (and lose all work) is welcomed;)
anyway, the way over rpc (4.6x) is the most elegant way..
but thanks for posting
Single WU cancellation is a
)
Single WU cancellation is a feature of 4.50 and later.
BOINC WIKI