martin f krafft wrote:
also sprach Blair L Strang <bls@xxxxxxxxxxxxxxxxxxxxx> [2005.03.15.2256 +0100]:
Sorry I didn't understand from your original post that this was
only happening occasionally. Duh!
It does only happen occassionally...
Perhaps look into ip_conntrack_max?
I don't have such a file. ip_conntrack_expect is the only other
one...
It /is/ a bit of a long shot because you probably would have noticed messages
saying "ip_conntrack: maximum limit of <n> entries exceeded" from your kernel.
But
worth a look anyway.
ip_conntrack_max is a sysctl which determines how many conntrack entries are
kept.
See: /proc/sys/net/ipv4/ip_conntrack_max.
Comparing this with "wc -l /proc/net/ip_conntrack" will tell you how close to
the
limit you are at a given point in time. The numbers can change pretty
dramatically
depending on use or abuse; a single nmap -sU -T Insane will chew through a lot
of
conntracks (1600 or so at peak when I tried it).
Ta,
Blair.
--
M-x yow!
Well, O.K. I'll compromise with my principles because of EXISTENTIAL DESPAIR!
| 