FIN_WAIT_2

[zurida]

How do I blow away the FIN_WAIT_2 for HPUX 10.20? Is there any script or command that I can run?

[malay boy]

I'm confuse...blown away...what do you mean ???


regards mB

[malay boy]

ok..sorry for previous reply.My brain a bit khok-khu this morning..I guest need a another coffee.

I found a good explanation from mr manoj as below :

Hi,

When the server closes a TCP connection, it sends a packet with the FIN bit sent to the client, which then responds with a packet with the ACK bit set. The client then sends a packet with the FIN bit set to the server, which responds with an ACK and the connection is closed. The state that the connection is in during the period between when the server gets the ACK from the client and the server gets the FIN from the client is known as FIN_WAIT_2. See the TCP RFC for the technical details of the state transitions.

you can get the value of FIN_WAIT_2 by using the following command:

ndd -get /dev/tcp tcp_fin_wait_2_timeout

and set it to a value let's say 11 minutes ..

by using
ndd -set /dev/tcp tcp_fin_wait_2_timeout 660000

OR modify
/etc/rc.config.d/nddconf

TRANSPORT_NAME[0]=tcp
NDD_NAME[0]=tcp_fin_wait_2_timeout
NDD_VALUE[0]=660000

This will force close them after 11 minutes..


I hope this helps ..

Manoj

http://forums.itrc.hp.com/cm/QuestionAnswer/1,,0xea7d972194d6d5118ff40090279cd0f9,00.html

[zurida]

Thanks Malay Boy.I'm using HPUX 10.20. I think the command ndd is for HPUX 11.X. I tried to man ndd using 10.20 but there is no manual on that. If u know the similar command for 10.20, is would be useful for me. Thanks.

[T G Manikandan]

On 10.20 you can use nettune

to list
nettune -l tcp_fin_wait_timer

to set the value--300 seconds

#nettune -s tcp_fin_wait_timer 300

Revert

[zurida]

Hi Manikandan,

I've tried using the command that u gave,
nettune -l tcp_fin_wait_timer
but I received this message
ioctl on `/dev/netman` failed (object name not found)Invalid arguement

Could u help. I really appreciate it.Thanks

[T G Manikandan]

nettune does not support the configuration of
tcp_fin_wait_timer unless the 10.20 ARPA patch is applied.

check for the patch
http://www1.itrc.hp.com/service/patch/patchDetail.do?patchid=PHNE_22507&context=hpux:800:10:20

check for the instructions of the patch.

This patch would require a REBOOT of the machine.

[Ron Kinner]

You might also look at what is causing these hangs. Do you have any Win NT boxes talking to this 10.20? They are notorious for causing this. There is an unreleased patch for tcpip.sys for NT which fixes the problem. You have to call MS support and tell them you need it.
http://support.microsoft.com/default.aspx?scid=kb;en-us;254930

They will waive the usual support fee or you can just download the patch from my website:

http://www.geocities.com/heartland/9670/ look at the bottom for tcpip patch.

Ron

[Paul Thomson_2]

Hi, we had this problem on 10.20 systems and we installed PHNE_19936 to clear the fin_wait_2.

We amended /sbin/init.d/net adding the line
/usr/contrib/bin/nettune -s tcp_fin_wait_timer 675
under the text
NETSTAT_DATA=/var/adm/netstat_data
So the file read
# We need /usr/sbin for lanconfig (for now)
export PATH=/sbin:/usr/sbin:$PATH
NETCONF=/etc/rc.config.d/netconf
NETSTAT_DATA=/var/adm/netstat_data
/usr/contrib/bin/nettune -s tcp_fin_wait_timer 675

You can check with nettune -l tcp_fin_wait_timer.

If you receive the /dev/netman message, try
mknod , an example of the file should be as
crw-rw-rw- 1 root other 60 0x000000 Jun 10 1996 /dev/netman

Iused the following script to install it automatically
(We had problems with getting downtime on 200 plus servers0 Hence the script. You may want to check for patches which supersede 19936 though, hope this helps.

Cheers

[Paul Thomson_2]

Just to add, we had this problem when the oracle listener shutdown and the port stayed in fin_wait_2.

Once we have installed this patch and config, the issue has reduced greatky, we still get the odd issue, but it is significantly less than we had before this solution.

[zurida]

Hi Guys,

Thanks for your help. I did a swlist. The lastest patch that we have on ARPA transport is PHNE_16210. I think the lastest patch would be PHNE_22507. I'll try to download the patch.
What is the command for the mknod if I have the message on netman?

[T G Manikandan]

you can use

# mknod /dev/netman c 60 0x000000

I would recommend loading the patch which will do things automatically


Thanks

[malay boy]

This example for mknode for creating a vg:

mknod /dev/vgnew/group c 64 0x030000

so ll will be :

crw-r--r-- 1 root sys 64 0x000000 May 13 2003 /dev/vgnew/group

I guest for your case you should specify 60 instaed of 64.And make sure that 0x030000 this number no body use before.

regards
mB

[rick jones]

keep in mind that the fin_wait_timeout setting is generally a massive kludge to work around broken applications.

In the case of applications not being able to restart, they need to be setting SO_REUSEADDR before trying to bind() thier well-known ports.

In the case of other applications, it is either their use of abortive close(), or their not setting SO_KEEPALIVE or some combination of the two.

ftp://ftp.cup.hp.com/dist/networking/briefs/annotated_nettune.txt