27 Apr
2001
27 Apr
'01
2:23 a.m.
You know, as I thought about it just switching off the power on the newer
unix boxes probably IS less destructive since the power button on newer
units is not actually tied directly in to the power supply. I am thinking
of newer RS6k's and HP9000 machines I have seen where the power switch
merely makes init do a powerfail shutdown (which is more graceful than
yanking the power cord or hitting the power on machine where the button is
actually connected to the PS)
We had our computer room UPS blow a capacitor and a non-electrician
facility person responding to the fire alarm cut the power to the room
(rather than let the make before break switch merely cutoff power to the
UPS) and our 400 user RS6k with an ISAM database needed two days of
database rebuilding before it was functional. The modern JFS filesystems
came back nice and quick but that did not stop the user data from being
corrupted.
Its replacement we are gradually cutting over to which uses a modern
transaction log based database was back up when the power was restored.
Likewise, on VMS without RMS journaling package installed you are likely
to get RMS-?-BUG, RMS-?-IRC, etc errors on machines with busy indexed
files switched off without a graceful shutdown.
On Fri, 27 Apr 2001, R. D. Davis wrote:
On Thu, 26 Apr 2001 healyzh(a)aracnet.com wrote:
--
Define older hardware. On my Sparc 10 and 20 I
run Solaris 2.6 as it's the
last version to support some of the hardware I use. Though I've been
Ah, you're using newer hardware. ;-) I was referring to a Sun 4/110
running Solaris 2.4... as well as a 3/60 that was running SunOS 4.1.1.
Both my SparcServer 1000 and my Ultra are running Solaris 8. Now, I'm
not in the habit of just switching off machines; I prefer to shut them
down properly whenever possible. However, if the machines go down due
to a power failure, etc... or just a test to see how they react to
being switched off before I begin to load software and data onto them
and use them for anything. The ones running Solaris 8 are much more
likely to necessitate a manual fsck than my other machines for some
reason; not sure why. Any ideas?
threatening to move the Sparc 20 to Solaris 8
when I upgrade its harddrives.
Out of curiosity, what hard drives are you using that Solaris 8
doesn't support? Will it support most older 5-1/4" SCSI hard drives
and SMD drives, or the Emulex SCSI <-> ESDI apater?
I'll put it another way, if you have the
chance to shut a UNIX box down
properly and don't, you're a fool. Especially if the system is like my one
I agree, which is why I make every attempt to do so. Not only do I
not want to risk losing data from a crash, but I also don't want to
waste my time having to to a restore that should never have been
necessitated in the first place.
Linux box and has a lot of disk space hanging off
of it (about 70GB), as
that's some serious fsck time! Personally if a system has an option to shut
it down gracefully, that's what I do, no matter the OS.
It's bad enough to have to wait for a check through 768MB of RAM,
without having to endure a manual fsck through quite a few gigabytes
of disk space.
Wow, you must really trust Linux! Your experiences with it are
probably different than mine, but I don't have enough faith in it,
based upon prior experiences with RedHat Linux. Based upon those
experiences, RedHat Linux appeared to have a less stable filesystem
than FreeBSD, and was more prone to need lots of help with a manual
fsck when power was lost, or when the machine crashed while running,
which wasn't uncommon---particularly if Netscape was being used on it.
What flavor of Linux are you using?