Date: Fri, 20 Feb 1998 13:42:06 -0500 (EST) From: "John T. Farmer" <jfarmer@goldsword.com> To: bad@uhf.wireless.net, bag@sinbin.demos.su Cc: agdolla@datanet.hu, freebsd-isp@FreeBSD.ORG, jfarmer@goldsword.com Subject: Re: fault tolerant :)) setup Message-ID: <199802201842.NAA23134@sabre.goldsword.com>
next in thread | raw e-mail | index | archive | help
On Fri, 20 Feb 1998 09:25:06 -0600 (CST) Bernie Doehner said: >What do you use on the backup sever to determine that the primary is dead? >I assume something like pinging the other host? Ping's not a good choice, too many variables that effect it's response. I would use a pair of monitoring programs that communicated directly with each other. Each program would have to "speak" to the other every time period, else it's assumed that "He's dead, Jim..." I would also consider by-passing the ethernet connection in favor of a serial or parallel port connection between the two. After all, one of the items your'e monitoring for potential failure is the NIC... >Is there any way you know of to set up a redudant system such that the ISA >cards from the primary sever are automaticaly hot switched upon dead of >the primary over to the secondary? Back in my "old days" circa 1980, we used what was called a DT07 switch unit with DEC PDP-11's, Vax's, etc. It was a programmable unit that plugged into two "Master" busses (Unibus) and moved a third "slave" bus between them. While I suspect that there might be such a beast available for the ISA bus, it begs the question, why bother? If you mean to use the PCI bus, then it's a specialized application of a PCI-PCI bridge, and it's a "simple matter of design..." (And Yes, I know what that entails...) In a server what's on the bus (PCI or ISA) that would need switching? Most servers, the only items to consider would be the NIC(s) and the disk adaptor(s). Terminal/access servers might have 1 or more port cards, Router/PC's the WAN card(s). In all these cases, there is a ton of state information that just switching the card/bus won't transfer unless you replicate the info to the secondary server AND can use it to startup a "sleeping" device. So to switch the cards over would require re-writing the selected ethernet card driver, SCSI card driver.... you get the picture. Far better to look at it at the external level where you are dealing with connecting/disconnecting the external ethernet & SCSI channels. For flexibility, I would approach from the basis of load sharing, with failure of one server being the extreme case. You can design for just about any level of "fault-tolerance" and availablity level that you want, but I ask my clients two questions when they start talking about it, 1.) "What levels do you want?" and 2.) "How big is your checkbook?" In other words, the closer to total fault-tolerance/recovery and 100% availability you get, the more zeros on the check you will give me... John ------------------------------------------------------------------------- John T. Farmer Proprietor, GoldSword Systems jfarmer@goldsword.com Public Internet Access in East Tennessee Office: (423)691-6498 for info, e-mail to info@goldsword.com Network Design, Internet Services & Servers, Consulting To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-isp" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199802201842.NAA23134>