SUMMARY: monitor and FC

From: <Eskil.Swahn_at_LDC.lu.se>
Date: Thu, 10 Aug 2000 11:39:51 +0200 (MEST)

Hi!

One response was enough. Thanks Alan. This solved the problem.

/Eskil

+--------------------------------------------------------------------+
| |
| Eskil Swahn |
| LDC, Lund University |
| Email: Eskil.Swahn_at_LDC.lu.se Web: http://www.ldc.lu.se/ |
| Phone: +46 46 222 4986 Fax: +46 46 138225 |
| |
+--------------------------------------------------------------------+

---------- Forwarded message ----------
Date: Wed, 09 Aug 100 10:35:13 -0600
From: alan_at_nabeth.cxo.dec.com
To: Eskil.Swahn_at_ldc.lu.se
Subject: Re: monitor and FC


        Edit the Makefile for the particular version and change
        the constant -DNEXUS_LIST to -DDI_AVQUE. Rebuild disk.c,
        queue.c and read_kernel.c, relink and see if it works
        better.

        The last gatekeeper and Freeware version of Monitor tried
        to traverse the internal SCSI driver data structures to
        get the depth of the command queue for each device. There
        have been cases where those data structure aren't well
        behaved to the outside world. I wouldn't be surprised if
        the Fibre Channel changes in V4 changed the structures
        or at least didn't maintain them as desired.

        By switching to "DI_AVQUE" to trust the queue length data
        from the table(2) data structure for disks, which seems
        to moderately reliable on V4 (not always, but better than
        a core dump).
Received on Thu Aug 10 2000 - 09:41:25 NZST

This archive was generated by hypermail 2.4.0 : Wed Nov 08 2023 - 11:53:41 NZDT