Re: SGI CRASH WHEN CONNECTED TO THE INTERNET

From: Alan Olson <awo_at_helix.nih.gov>
Date: Fri, 10 Oct 2003 11:52:24 -0400

>Date: Fri, 10 Oct 2003 09:22:24 -0400
>From: Eric Gallagher <gallaghe_at_ninds.nih.gov>
>X-Accept-Language: en-us, en
>To: Alan Olson <awo_at_helix.nih.gov>
>Cc: Doug Morris <morrisd_at_ninds.nih.gov>,
> "Martin (NINDS)" <LizakM_at_ninds.nih.gov>
>Subject: Re: Fwd: SGI CRASH WHEN CONNECTED TO THE INTERNET
>X-Sanitizer: murus Advosys mail filter
>
>Dear Zhang,
>
>This is a known problem in dual-hosted SGI O2s.
>
>One thing that may help is rebooting the O2 into the command monitor
>mode, which is an option when you interrupt the normal bootup. From
>the command monitor > prompt, hard code the network speed and duplex
>settings for the NIC in question, as in
>
> > setenv -p ec1mode f100
> > setenv -p ec1mode h100
> > setenv -p ec1mode f10
> > setenv -p ec1mode h10
>
>In most single-hosted cases, the network setting would involve ec0.
>But in our O2 instruments, the syntax is as above. The one that
>we've had the most luck with is
>
> > setenv -p ec1mode h100
>
>which sets the duplex mode to half-duplex at 100 mbps. It may be
>necessary for your friend to set the speed even lower, to h10.
>Definitely do *not* try to do full duplex. The 02 bus chokes when
>it tries to handle two full-duplex transmissions, the other one
>being the data acquisition.
>
>As I remember it, we (Marty and I) also went into the kernel (via
>systune) to fix the ecf_max_rxds problem but to almost no effect.
>Increasing the buffer available did not stop the errors. The only
>real, if tenuous, solution was to stop the NIC ec1 auto-negotiation
>and crank it back to half duplex. Marty may chime in and add details
>I've missed.
>
> Eric
>
>
>Alan Olson wrote:
>
>>Eric, Doug, Marty -
>> Any insights for this guy?
>>
>> - Alan O.
>>
>>>From: "Zhang, Weixing" <Weixing.Zhang_at_stjude.org>
>>>To: "AMMRL (E-mail)" <ammrl_at_chemnmr.colorado.edu>
>>>Date: Thu, 9 Oct 2003 15:42:22 -0500
>>>Subject: SGI CRASH WHEN CONNECTED TO THE INTERNET
>>>X-OriginalArrivalTime: 09 Oct 2003 20:42:22.0389 (UTC)
>>>FILETIME=[D6F8E650:01C38EA5]
>>>X-SEF-723560E2-3392-41F8-A983-A3F5486E94A: 1
>>>Thread-Topic: SGI_CRASH and 2nd Hard-drive on Linux box
>>>Thread-Index: AcOMRovIhQiM+BOHSamcHeyBtO6lYgCXjRCg
>>>X-Sanitizer: murus Advosys mail filter
>>>
>>>Hello:
>>>
>>>The SGI O2 (IRIX 6.5, xwinnmr 3.5) often crashes when it is connected to the
>>>internet. If it is not connected to the internet, everything is OK.
>>>The ethernet for network connection is ec1.
>>>The ethernet for BRUKER NMR console (spect) is ec0.
>>>
>>>> I have checked the /var/adm/SYSLOG file for error messages
>>>
>>> and found the following (repeating) error messages:
>>>
>>>> Jul 29 07:02:37 1A:chamlang unix: |$(0xc4)ALERT: ec1: need more
>>>>ecf_max_rxds (7)
>>>> Jul 29 07:27:19 4A:chamlang unix: |$(0xc3)WARNING: ec1:
>>>>statistics overflow
>>>> Jul 29 07:27:19 4A:chamlang last message repeated 4 times
>>>
>>> .....................
>>>
>>>> Oct 6 20:46:05 1A:chamlang unix: |$(0xc4)ALERT: ec1: need more
>>>>ecf_max_rxds (2)
>>>> Oct 6 20:52:39 4A:chamlang unix: |$(0xc3)WARNING: ec1:
>>>>statistics overflow
>>>> Oct 6 20:52:39 4A:chamlang last message repeated 4 times
>>>>
>>>> What should we do?
>>>>
>>>> Thanks,
>>>>
>>>> Weixing Zhang,
>>>
>>
>
>--
>Eric Gallagher Senior Systems Analyst
>NIH/NINDS/NMRF phone: 301-402-3110
>geric_at_nih.gov fax: 301-402-0119
Received on Fri Oct 10 2003 - 18:14:41 MST