When Apache gets too many 408 errors at the
beginning of its log file, awstats complains about the log
file not being in the correct format and will not proceed
until the leading 408 errors are removed. 408 errors later
in the log file are processed (apparently) as corrupted
lines. The 408 error lines are of the following format:
xxx.xxx.xxx.xxx - - [01/Jun/2003:04:49:08 -0400] "-"
408 - "-" "-
This is in the correct format (LogFormat=1, Apache
combined format), but since there is no real data it
errors out. This 408 traffic is conjectured to be caused
by CodeRed attempts that are detected by firewalls or
routers and have their payloads dropped (see
http://cert.uni-
stuttgart.de/archive/incidents/2002/02/msg00015.html).
I'm not exactly sure this would be a bug or just
something that needs to be dealt with by the end-user of
awstats, but I wanted to give a heads-up anyways.
Excellent software, BTW...
Thanks.
Logged In: YES
user_id=96898
to avoid this you can increase parameter
NbOfLinesForCorruptedLog=50
to
NbOfLinesForCorruptedLog=10000
for example.
Just FYI, this is still happening...
... and it was only 2 lines in the whole log!
Last edit: Jan Grewe 2012-11-20
This is still happening.
I don't think that having a really large number in
NbOfLinesForCorruptedLogwould ever be a good solution.You should be able to ignore 408 entries. Actually, as a stat tool, you should count them and show them in our reports.
Especially, newer versions of Apache now have better handling of the slowloris attacks and will generate more 408 errors as a result.