On Tue, 28 Aug 2001, Per Jessen wrote:
Hhmmm, that's just gratuitous eye candy, no? Besides, it's pretty trivial to replace those graphics with logos or other stuff. I think that doing things that are new and usefull (like serial consoles) would really help get people and hardware manufacturers interested. Imagine being able to run a server farm with no graphics cards, just a serial console and ethernet...
we make it easier. We do all our admin on linuxbios clusters over ethernet. Serial console admin is not our idea of a good time.
To buy, see www.linuxlabs.com
ron
- To unsubscribe: send mail to majordomo@freiburg.linux.de with 'unsubscribe openbios' in the body of the message
On Tue, 28 Aug 2001, Ronald G Minnich wrote:
On Tue, 28 Aug 2001, Per Jessen wrote:
Hhmmm, that's just gratuitous eye candy, no? Besides, it's pretty trivial to replace those graphics with logos or other stuff. I think that doing things that are new and usefull (like serial consoles) would really help get people and hardware manufacturers interested. Imagine being able to run a server farm with no graphics cards, just a serial console and ethernet...
we make it easier. We do all our admin on linuxbios clusters over ethernet. Serial console admin is not our idea of a good time.
To buy, see www.linuxlabs.com
Right, but what happens when the switch barfs @ 4am 500miles away? With a serial console, I can hookup either a modem or a portmaster. If the network fails, I can call the data center, tell them "turn on modem in cage 12" and dialup do fix whatever is wrong (BTDT). That's one of the reasons why Sun hardware is so popular in server farms.
Sure, ethernet is the way I admin servers most of the time, but it does not help when:
a. the network is down, either lan or wan b. trying to see boot msgs for failed hardware
Unless, of course, you have a special ethernet driver that is embedded in the bios and you can connect to at first hardware init.
I seem to recall that you developed linuxbios specifically for use in Beowolf clusters, not with the intent of having geografically remote systems.
I could be totally wrong, tho'... BTW, some chipsets include a serial console, which is great.
Chris.
-- chris maresca internet systems architect -- www.chrismaresca.com
"linux, only up 138 days, because california has flaky power... "
- To unsubscribe: send mail to majordomo@freiburg.linux.de with 'unsubscribe openbios' in the body of the message
On Tue, 28 Aug 2001, Chris Maresca wrote:
Right, but what happens when the switch barfs @ 4am 500miles away? With a serial console, I can hookup either a modem or a portmaster. If the network fails, I can call the data center, tell them "turn on modem in cage 12" and dialup do fix whatever is wrong (BTDT). That's one of the reasons why Sun hardware is so popular in server farms.
our experience is different. We've had way more trouble with serial port concentrators, cables etc. than enet switches. Overall the scenario you describe would not work well for us. It does work for many people however.
b. trying to see boot msgs for failed hardware
This is an important one. You're right here. But rather than go with the serial port concentrators we're probably going to put console messages out over enet.
I seem to recall that you developed linuxbios specifically for use in Beowolf clusters, not with the intent of having geografically remote systems.
you're right. We can't address your problems (yet).
ron
- To unsubscribe: send mail to majordomo@freiburg.linux.de with 'unsubscribe openbios' in the body of the message
On Tue, 28 Aug 2001, Ronald G Minnich wrote:
On Tue, 28 Aug 2001, Chris Maresca wrote:
Right, but what happens when the switch barfs @ 4am 500miles away? With a serial console, I can hookup either a modem or a portmaster. If the network fails, I can call the data center, tell them "turn on modem in cage 12" and dialup do fix whatever is wrong (BTDT). That's one of the reasons why Sun hardware is so popular in server farms.
our experience is different. We've had way more trouble with serial port concentrators, cables etc. than enet switches. Overall the scenario you describe would not work well for us. It does work for many people however.
Hhhmm, unfortunetly that has not been my experience. The problem is not usually the switch, but the WAN. A site's network gets knocked over (think Yahoo's recent denial of service attack or some such), and there is no longer a way to access the machine remotely. It's in Austin, Seattle or NYC and I'm in San Francisco. Even with backend networks, it's still a huge problem. Portmasters and their ilk are very reliable once setup (but they are a b*tch to setup correctly) as are modems and phone lines.
Even when you might have a dialup to a machine on the same segment, it's possible that the machine has been so overload/damaged/whatever that it no longer responds in a meaningfull way to network connections. However, serial connection almost always works.
Note that this is the way that most large remote datacenters I have seen are setup, I would think that a serial console would be very wanted and usefull. As others have pointed out, there are companies that take advantage of this by producing products that mimick a serial console.
Another usefull and similar addition would be a BIOS based watchdog program, but that's another topic...
Chris.
-- chris maresca internet systems architect -- www.chrismaresca.com
"linux, only up 138 days, because california has flaky power... "
- To unsubscribe: send mail to majordomo@freiburg.linux.de with 'unsubscribe openbios' in the body of the message
-
Chris Maresca -
Ronald G Minnich