Subject: | |
From: | |
Reply To: | |
Date: | Wed, 22 Jan 1997 00:12:37 +0200 |
Content-Type: | text/plain |
Parts/Attachments: |
|
|
Thanks, Jeff
I'm going to bed now, but I will pursue this tomorrow.
BTW, the system will be upgraded to 5.5 on Saturday. Is it fixed in 5.5?
Regards
Neil
>----------
>From: Jeff Kell[SMTP:[log in to unmask]]
>Sent: 22 January 1997 12:00
>To: Neil Harvey
>Subject: Re: [HP3000-L] Cobol and Sockets - longish!
>
>Neil Harvey wrote:
>> It that right - Max 256, used 131, but sockinfo says only 16?
>> Tell me I'm not mad!!!!
>
>You're not mad, as in crazy; but you might be mad, as in angry!
>
>Yes, you are losing sockets to the old state 0.0 or 0.c status; the more
>common definition is the "Fin-Wait-2" tcp connection status.
>
>I wrote this up in my httpd document:
> http://opus.admin.utc.edu/http3000.htm
>
>Specifically:
>Network transport may hang after several days of continuous network
>operation and/or a low number of configured TCP connections in NMMGR's
>NETXPORT.GPROT.TCP screen. This problem manifests itself by refusing
>connections and failing even local NS services with a "transport not
>started" error. None of the "usual" tools show any sign of a problem,
>but if you take a system dump, all available TCP connections will be
>"hung". Refer to SR 4701-308304, a beta patch is available NSTED54.
>This problem is either unique to, or at least very sensitive to httpd.
>
>You probably want an even later patch as I think there have been fixes
>since then.
>
>Jeff Kell <[log in to unmask]>
>
|
|
|