hot swap SCSI SCA bus transients harmful to non SCA drives?
 

hotswapping scsi sca is as easy as

echo "scsi add-single-device 0 0 2 0" > /proc/scsi/scsi
echo "scsi remove-single-device 0 0 2 0" > /proc/scsi/scsi

i believe SCA prescribes a low connector capacitance and some bus transient tolerance (low pass filter) for each device. if so, is it harmful to routinely hotswap SCA devices on a bus that also has non-SCA devices that may be more sensitive to transients?

TIA, -bp

Are we talking about non-SCA devices that use a physical adapter plate/plug?
ie: if you're using a 68->80 pin adapter it should present itself (electrical/mechanical/signal) to the SCSI bus the same as a native SCA drive.

Also curious about the phrase "routinely hotswap" - I don't think the physiacal interface is engineered for daily swaps.

hey mcleodnine, thx for the reply.

>non-SCA devices that use a physical adapter plate/plug?
i put four SCA 80-pin drives in scsi hotswap cradles that have 68-pin external interfaces.

the cradles are on the same bus as other 68-pin scsi drives that are not hotswapped nor hotswappable.

my concern is that hotswapping the sca drives might corrupt data or shorten the lifespan of 68-pin drives on the same bus since i do not believe that the same high-frequency (noise) filtering is specified for standard 68-pin u320 in comparison to SCA. i developed this notion by routinely finding references to spike-filtering in SCA specs but never in non-SCA specs.

having never fiddled with SCA before, i wonder if i'm just asking for trouble by mixing SCA with standard 68-pin drives.

as for interface robustness, each drive would be plugged in/out once every 8 days as i'm trying to store data offline in a chronological cascade in effort to minimize volatility. does this frequency sound unusually high, this seems like the kind of thing one learns (the hard way) by experience, which is something i surely don't have in this case.

tia, - bp

Ahhh. I was making the assumptoin you were plugging 68-pin drives into an 80-pin bus (probably reading my own experience into the situation).

My best guess (heavy on the guess here) is that the signal isolation for SCA is required to meet bus reliability specs for the higher-bandwidth it's engineered for. Although I can't say for sure I'd reckon you would be okay on the data front as all the drives on the bus will be operating at the lowest common speed present. SCSI hot-swap is a mature technology (seems reliable enough on the old SCSI-2 drives in the old Proliant RAID array)

As for the durability of swap-outs I have my doubts about the durability. You could proably dig throuh a stack of whitepapers to prove one way or another. One thing I will offer is that high-speed SCSI drives will happily cruise along at constant rotation and temperature - you'll usually see a failure after a long power-down event or (as in your case) hibernating drives once a week.

I'd like to hear if anybody can offer some more insight into this.

i noticed my IDE data drives stopped failing so often after i stopped using HDPARM to spin them down after a few hours of nonuse. i concluded frequent spinup/down was wearing them out at least twice as fast as comparable drives that spun for all time but i never compared this ratio with others' experience. most IDE drives were western digital special editon during that time.

>you'll usually see a failure after a long power-down event or (as in your case) hibernating drives once a week.
i wonder if this results from inductance forcing high current similar to lightbulbs typically failing when first turning on, i never considered that before. that might explain what i said above.

thx, -bp