w_tom said:
Doing something is just fine only because it did not cause
failures previously? What kind of logic is that?
I originally raised the question of whether hot-plugging standard PCI
cards has actually been known to cause damage, so had better point out
that I have always been careful to pointout that I don't advocate the
even if it doesn't actually appear to cause damage; and nobody posting
here has implied that hot-plugging is fine.
Some
darkening of the first O'ring was the only apparent effect
when launching a space shuttle in sub freezing weather...
Funnily enough, I had thought of the exact same example in this context.
An O-ring which was not getting eroded at all was getting eroded to one-
third of its thickness, so the project managers said that there was a 3-
to-1 safety factor, rather than that something unexpected was happening,
See Richard Feynman's report. Similar in principle, though not in
consequences, to hot-plugging, getting away with it, and adopting it as
standard practice.
The second fatal shuttle disaster was due to exactly the same attitude:
bits of foam come off and bash into the structure, though they're not
supposed to; we've got away with every time; so let's keep doing it.
I have simply asked whether damage has been observed without giving any
reasons. My reasons for asking the question are in fact:
a. Simple curiosity: is this actually seen to happen in practice?
b. If hot-plugging is done accidentally, what are likely to be the
consequences?
c. If there are legitimate reasons for hot-plugging: how dangerous
actually is it to motherboard and card?
A legitimate reason: an Adaptec 2940U2W PCI SCSI card has its BIOS
flashed unsuccessfully. The card is not recognised by the operating
system, and hence can't be re-flashed (in some case the machine will not
boot); the BIOS is soldered in; and the manufacturer cannot repair. The
card (fairly expensive to replace) is junk.
Possible attempt to repair: boot the machine with a good card, or with
no card; hot-unplug good card; hot-plug bad card; try running flash
procedure.
This issue has been discussed on Usenet, and hot-plugging has been done
and not caused damage. I have Googled for, but not found, instances of
damage.
[detailed comments on why not to hot-swap]
If these comments are for my benefit rather than a contribution to the
general discussion (as I raised the question) don't feel you need to
explain: I have designed and repaired circuitry and know and play by the
rules of the game, even if I ask questions about them.
As I keep on saying: I don't disagree with you in any way, I would just
be interested to hear about any actual observed consequences (or non-
consequences) of hot-plugging PCI. Even if it is usually harmless, there
are very few reasons to hot-plug in a machine which doesn't need to run
with zero downtime -- it would at best save a bit of rebooting time.
Best wishes,
