What's a Corrupted Fstat-file ?
13 May 2006 21:07:37 UTC
Topic 191221
(moderation:
I got it with this wu that i had to abort (with some others :( )
(the wu was so slow and didn't go on)
2006-05-13 02:09:05.8099 [CRITICAL]: Corrupted Fstat-file '../../projects/einstein.phys.uwm.edu/z1_1384.0__1664_S4R2a_0_0': has 0 bytes instead of 4270468.
[
What's a Corrupted Fstat-file ?
)
It's a 'working' file that keeps track of the state of the computation via checkpoints. For some reason yours was zero length for that work unit, that is empty! The computation was restarted from the beginning..... To further analyse for you I'd look up the error code in the Wiki, but it seems to be down right now.
Cheers, Mike.
(edit) There is the curious point of why 'Fstat file reached MaxFileSizeKB ==> compactifying ... done' if it was empty!
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
Well, no hits on the
)
Well, no hits on the Wiki..... :-(
Cheers, Mike.
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
Thanks Mike :) So the
)
Thanks Mike :)
So the Fstat-file was empty but it has been able to reach the MaxFileSizeKB
Ehm, amazing "grin"
[
The "Fstat.out" file is the
)
The "Fstat.out" file is the file that keeps track of intermediate results, the "Fstat.out.ckp" ist the checkpoint file that keeps some metainformation, including the expected lenght of the Fstat file "Fstat.out".
When the App is (re-)started, it first reads the checkpoint file (if present), then tries to read in the Fstat file with the information it got. If that works, the data is read into an internal structure that is called the "toplist", as you can see in the second block of stderr output. If the Fstat file doesn't agree with the information in the checkpoint file, it is considered to be corrupt, and calculation starts from the beginning.
I suspect that somehow a checkpoint file was left in the slots directory from a previous run, as this situation happened right at the beginning of the calculation. The Fstat file indeed is empty at the beginning of the calculation, which is not an error.
To save time in FileIO, new (intermediate) results get only appended to the Fstat file instead of rewriting the whole file every time. To not clobber the harddisk, however, when this grows too large, it is "compactified", i.e. newly written as a whole.
BM
BM