strange ip_conntrack values!
Wed, 2 Jan 2002 13:21:53 +0545 (NPT)
Further to my last mail...
I have observed the following:
grep conn /proc/slabinfo = 4756
grep ESTABLISHED /proc/net/ip_conntrack = 559
grep CLOSE /proc/net/ip_conntrack = 857
grep CLOSE_WAIT /proc/net/ip_conntrack = 200
grep TIME_WAIT /proc/net/ip_conntrack = 543
As you can see that CLOSE is taking the most no. of entries in the conn
db due to the HIGH timeouts. I also observed that CLOSE was the only one
with those HIGH timeouts and the rest had values within the given values.
Most CLOSE timeouts were 5/6/7 digit values (days!)...default is 10 secs.
Due to this strange behaviour my total conntrack entries will always
be higher than normal, right!
I am awaiting responses...I feel this is definitely a BUG now!
On Tue, 1 Jan 2002, Don Lafontaine wrote:
> Yes indeed, I've noticed this for quite some time.
> Here's a script I found that will display the table in a 'top' like fashion.
> ----- Original Message -----
> From: "Raj" <firstname.lastname@example.org>
> To: "Netfilter Mailing List" <email@example.com>
> Sent: Tuesday, January 01, 2002 7:26 AM
> Subject: strange ip_conntrack values!
> > HI,
> > I just noticed in my /proc/net/ip_conntrack file that the CLOSE timeouts
> > are shockingly high than the real value of 10 secs.
> > How could this be possible?
> > # cat ip_ct_tcp_timeout_close
> > 1000
> > tcp 6 1797884 CLOSE src=202.52.X.X dst=202.52.X.X sport=1026
> > dport=1 10 src=202.52.X.X dst=202.52.X.X sport=110 dport=1026
> > [ASSURED] use=1
> > tcp 6 105137 CLOSE src=212.45.X.X dst=202.52.X.X sport=1121
> > dport=80 [UNREPLIED] src=202.52.X.X dst=212.45.X.X sport=80 dport=1121
> > use=1
> > Thanks,
> > Raj