From owner-freebsd-stable Thu Sep 14 10:14: 7 2000 Delivered-To: freebsd-stable@freebsd.org Received: from yellow.rahul.net (yellow.rahul.net [192.160.13.18]) by hub.freebsd.org (Postfix) with ESMTP id B985337B423 for ; Thu, 14 Sep 2000 10:14:05 -0700 (PDT) Received: by yellow.rahul.net (Postfix, from userid 104) id 655227C3D; Thu, 14 Sep 2000 10:14:05 -0700 (PDT) To: freebsd-stable@freebsd.org Subject: Re: SCSI retries without errors in /var/log/messages? Newsgroups: a2i.lists.freebsd-stable References: X-Newsreader: NN version 6.5.6 (NOV) Message-Id: <20000914171405.655227C3D@yellow.rahul.net> Date: Thu, 14 Sep 2000 10:14:05 -0700 (PDT) From: dhesi@rahul.net (Rahul Dhesi) Sender: owner-freebsd-stable@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG jean-francois.dockes@wanadoo.fr writes: >I think that most SCSI retries should be logged. >Except in some specific cases of retries after a unit attention >condition, retries are usually indicative of hardware trouble. SunOS deals with soft memory errors in a very nice way. After a certain number of soft memory errors have occurred, it syslog's a message saying essentially: XXX corrected memory errors on memory chip YYY where XXX is how many times an error was corrected and YYY is the location of the chip on the motherboard. I think this is a very nice strategy. It avoids too many false warnings but still alerts the operator to consider replacing an unreliable memory chip. The same strategy could be used in any situation where correctable errors are occurring: Simply keep count of them and log a warning when a threshold is reached. You would, of course, also want to format the error message such that it is self-explanatory to the user who has not read the device driver source code. -- Rahul To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message