Article problems after rebooting server

Bill Sellers w.a.sellers at larc.nasa.gov
Fri Nov 18 22:08:04 UTC 2005


Just for the record, we took a power hit last week causing this error...

expireover[11284]: [ID 871830 news.error] CNFS-sm: cnfs_retrieve: token 
@03004230093C4E706D646E5668346D644232@: bogus cycbuff name: B0 
<Npmd:0xdcacd06800:1835287090

This caused innd to crash.  There is a valid cycbuff called B01, but not 
B0... Setting cnfscheckfudgesize worked like a charm.  Everything else I 
tried failed.  Is there something we can do to make innd more robust when 
it sees a cnfs problem?

Thanks,
Bill


At 04:28 PM 11/17/2005, you wrote:
>Francisco Cosin Vallejo <francisco.cosin at corp.ya.com> writes:
>
> > Yesterday I had to reboot server cause of high load left server KO.
>
> > Since then I have lost many messages due to caducity and now I'm getting
> >   thousands of errors like those:
>
> > Nov 17 10:15:45 news nnrpd[22865]: CNFS-sm: could not mmap token
> > @0300455342494E313000000CF59D00000073@ ESBIN10:0x19eb3a0c:115: Cannot
> > allocate memory
>
>You have a corrupt article in CNFS, probably with an irrationally large
>size.  Setting cnfscheckfudgesize (see inn.conf(5)) will cause CNFS to
>diagnose this earlier than the mmap failure, but either way that article
>is probably lost.  Eventually the cycbuff will roll over it again.
>
>--
>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.



More information about the inn-workers mailing list