CNFS offset problems

[Russ Allbery]

Bettina Fink <laura at hydrophil.de> writes:
> Jeffrey M. Vinocur <jeff at litech.org> wrote on December 29th:

>> CNFS problems started suddenly yesterday.  Any ideas?

>> Unknown entries from news log file:
>> First 50 / 1192 lines (4.2%)
>> Dec 28 18:51:56 puck innd: CNFS-sm: CNFSUsedBlock: invalid offset 9c4200, 
>> min = 4000, max = 9c4000
>> Dec 28 18:51:56 puck innfeed[5557]: CNFS-sm: could not match article size 
>> token @0301434F4E54524C310000004E1F0000006E@ CONTRL1:0x9c3e00:110: 1520
>> Dec 28 18:51:56 puck innd: CNFS-sm: CNFSUsedBlock: invalid offset 9c4400, 
>> min = 4000, max = 9c4000
>> Dec 28 18:51:56 puck innd: CNFS-sm: CNFSUsedBlock: invalid offset 9c4600, 
>> min = 4000, max = 9c4000
>> Dec 28 18:51:56 puck innd: CNFS-sm: CNFSUsedBlock: invalid offset 9c4800, 
>> min = 4000, max = 9c4000
>> Dec 28 18:51:56 puck innd: CNFS-sm: CNFSUsedBlock: invalid offset 9c4a00, 
>> min = 4000, max = 9c4000
>> Dec 28 18:51:56 puck innfeed[5557]: CNFS-sm: CNFSUsedBlock: invalid offset
>> 9c4400, min = 4000, max = 9c4000
>> [...]

> The same happened to one of my test boxes:

Yeah, I just had it happen to my test system too.  It's like it sometimes
loses track of where the end of the cycbuff is.  I have four disk files
used as cycbuffs, each 1GB in length, and during the second cycle on the
fourth one it just kept writing to it until it reached 2GB and died.

Are either of you using SEQUENTIAL?  And are you using disk files or block
devices?

-- 
Russ Allbery (rra at stanford.edu)             <http://www.eyrie.org/~eagle/>

    Please send questions to the list rather than mailing me directly.
     <http://www.eyrie.org/~eagle/faqs/questions.html> explains why.