RE: Unstable Master

[Dale Scheetz <dwarf@polaris.net>]

Simon, sorry to have pulled your chain so hard. It was not my intention.

This problem seems to be more complex than just the scsi problem.

On Thu, 25 Apr 1996, Simon Shapiro wrote:

> 
> Hi Dale Scheetz;  On 25-Apr-96 you wrote: 
> > > Yesterday new-master.debian.org was unavailable most of the day. This
> > morning I was able to get in, but had a lockup before downloading more
> > than 5 files. New-master.debian.org seems to have gone off the air, again.
> 
My experience was no contact via ping for most of the day. On the rare
times that ping got a response I was still not able to establish ftp
connection.
I have been told that the DNS problems have been solved, but just before
beginning this post I pinged both master and new-master:

PING master.debian.org (192.6.120.248): 56 data bytes
-----------  no response  --------------------

PING new-master.debian.org (206.139.73.161): 56 data bytes

got the last three of six pings sent, average time 330 ms

traceroute looks pretty ugly...

 1  portal2.polaris.net (199.44.34.11)  144.601 ms  141.516 ms  139.636 ms
 2  199.44.34.1 (199.44.34.1)  139.571 ms  129.492 ms  139.632 ms
 3  199.44.33.13 (199.44.33.13)  159.568 ms  149.528 ms  139.752 ms
 4  cisco.cntfl.com (199.44.1.1)  139.461 ms  199.485 ms  239.671 ms
 5  tlh877_hs_r3.cntfl.com (199.44.9.35)  149.546 ms  159.485 ms  149.686 ms
 6  sl-fw-14-H12/0-T3.sprintlink.net (144.228.134.9)  179.535 ms  179.524 ms  169.581 ms
 7  sl-fw-6-F0/0.sprintlink.net (144.228.30.6)  239.538 ms  219.463 ms  249.638 ms
 8  sl-ana-1-H2/0-T3.sprintlink.net (144.228.10.30)  269.578 ms  189.51 ms  189.628 ms
 9  sl-ana-2-F0/0.sprintlink.net (144.228.70.2)  359.584 ms  369.473 ms  239.661 ms
10  sl-stk-6-H2/0-T3.sprintlink.net (144.228.10.25)  259.597 ms  209.52 ms  199.636 ms
11  sl-mae-w-H1/0-T3.sprintlink.net (144.228.10.46)  219.588 ms  219.557 ms  209.662 ms
12  san-jose3.ca.alter.net (198.32.136.42)  239.712 ms  219.148 ms  219.582 ms
13  * Hssi2/0.CR1.SCL1.Alter.Net (137.39.100.2)  220.228 ms  219.385 ms
14  107.Hssi4/0.CR1.CHI1.Alter.Net (137.39.30.169)  269.53 ms * *
15  Fddi0/0.CR2.CHI1.Alter.Net (137.39.35.66)  270.139 ms  249.464 ms *
16  Hssi3/0.Chicago2.IL.Alter.Net (137.39.100.82)  260.105 ms  259.439 ms  269.644 ms
17  * Fddi1/0.Chicago1.IL.Alter.Net (137.39.137.1)  260.243 ms  349.429 ms
18  Serial1/7.Portland1.OR.Alter.Net (137.39.101.70)  329.548 ms * *
19  * iconnect.ALTER.NET (137.39.218.82)  340.267 ms  329.332 ms
20  new-master.debian.org (206.139.73.161)  349.527 ms *  350.229 ms


> Most ofthe day?  I was here from 0845 until 1900 PDT and it was working.  
> I dis take it down for about 10 minutes, twice to re-boot a new kernel, 
> after Matt told me he was caught up.
> 
> Indeed this morning it was dead.  Panicked again.  Please understand that
> I do not have SCSI problems.  Linux kernel 1.3.whatever does.  It is broken
> and buggy.  Namely, it will crash under load of certain types.  I have four
> Linux machines running here.  ALL will panic on SCSI unless you avoid certain
> things.  These operations cannot be avoided on master :-(
> 
> I am spending much too much time on tracking this problem. 
> 
> > Is Simon still having scsi problems with the buslogic card?
> 
> The problem is not with the Buslogic cards.  It is in what is called ``scsi
> midlevel code'', centered around drivers/scsi/scsi.c.  It has such an 
> architecture as to guarantee that command queues WILL overrun, and when 
> they do overrun, there will be no way to recover.  Hence, panic.
>
My mistake. I thought it was in the buslogic driver. Being in the midlevel
driver makes the problem more general than a specific driver.
 
> > It would be very helpful to the rest of us if Simon could advertise the
> > time windows when testing new kernels will NOT be going on at master.
> 
> I will, to the debian-masters list.
> 
> > Previously he suggested that evenings would be more stable, but my
> > experience lately has been less than positive.
> > Hope we get this resolved soon...
> 
> I am running 1.3.77+ (specially patched 1.3.77) on master.  I did re-boot
> 1.3.90 for two minutes a few days ago.  Then I tried 1.3.94 yesterday for about
> 2 hours.  Every time i do that, I first try the kernel on another host first.
> If it boots cleanly and stays up for a day or so, I try nomis, then master.
> 
> I hope I can get over these soon.  I take no joy in crashes, nor in taking mail
> such as this.  We decided to run master on Linux kernel.  Until such kernel 
> stabilizes, we will have a yoyo.  We can always setup an NT ftp server.  It cras
> hes
> with CC:mail but we do not care.   Right?
> 
> If any of you want to participate in a re-design of hte Linux SCSI subsystem,
> drop me a note.
> 
> And PLEASE, PLEASE, please:
> 
>   If you witness disruption of service on master, please contact me in the order
>   below:
> 
>   0.  Mail help@i-Connect.Net
>   1.  Mail shimon@i-Connect.net
>   2.  Try ``talk shimon@i-Connect.Net''
>   3.  Call 503.677.2900
>   4.  Call 503.677.2898
>   5.  Dial 503.441.1681 and leave 1111 as a paging no.
> 
> If none of these worked, it means I am dead, nobody mans the support line, nobod
> y is in the
> office and our entire feed is dead and there is no telephone service in Beaveron
> , OR.
> 
> 
> Sincerely Yours,           (Sent on 04/25/96, 14:35:48 by XF-Mail)
> 
> Simon Shapiro
> Director of Technology     i-Connect.Net, a Division of iConnect Corp.
> Shimon@i-Connect.Net       13455 SW Allen Blvd., Suite 140 Beaverton OR 97008
> 
> 

Thanks for all the info,

Dwarf

------------                                          --------------

aka   Dale Scheetz                   Phone:   1 (904) 877-0257
      Flexible Software              Fax:     NONE 
      Black Creek Critters           e-mail:  dwarf@polaris.net

------------ If you don't see what you want, just ask --------------