CNFS mmap memory error
Antoine Delvaux
antoine.delvaux at belnet.be
Wed Mar 20 17:39:38 UTC 2002
Le 2002.03.20 01:39, Katsuhiro Kondou a =E9crit :
>=20
> In article <20020319153251.M15867 at aegir.fw.belnet.be>,
> Antoine Delvaux <antoine.delvaux at belnet.be> wrote;
>=20
> } Mar 19 15:15:07 aotearoa innfeed[1036]: CNFS-sm: could not mmap token
> } @03015344453700000000001AC32B00000007@ SDE7:0x3586560c:7: Cannot
> } allocate memory
> } Mar 19 15:15:07 aotearoa innfeed[1036]: Could not retrieve
> } @03015344453700000000001AC32B00000007@: mmap failed
> }
> } Does this means that my buffers are to big (maybe just a bit more than
> } 2 GB) ?
>=20
> No. cnfs has a header for each article. It includes
> the size of article, but it doesn't look like your
> header is what it should be. There are three possibilities
> how this comes.
>=20
> - the buffer is broken(this may happen if the system was
> crashed)
This a system restarted from scratch 24 hours ago... CNFS were empty
> - new article is stored into buffer where innfeed is
> trying to retrieve
How can this happen ? I've got 104 GB of spool and the cycle time seems=20
to be about 7h.
In the logs the token seems to be always nearly the same (or not varying=20
much at last), is this token the cnfs header of the article ? What does=20
it means if the token is always nearly the same ? Exemples :
@0301423531000000000000139E4700000002@
@030142343400000000000014107300000002@
@03014230370000000000001359F700000002@
@030142303800000000000014B5C200000002@
@0301423138000000000000145D7A00000002@
^ ^ ^^^^^
Only these 7 digits do change in the log file for more than 20 hours=20
(which seems incoherent with the buffer cycle time...). The buffer and=20
the address to map seems to vary a lot though.
I think I don't understand all of these cnfs buffer stuff much.
Antoine.
--=20
http://www.belnet.be Tel: + 32 2 790 33 33
BELNET Service Support Team Fax: + 32 2 790 33 34
More information about the inn-workers
mailing list