<p>I think you're right on chipset and determinism. <br>
I would almost bet that a hotswap move then reboot would rename drives in bus/port order. So pull sdc and replace to be sdg will on reboot become sdc again.</p>
<div class="gmail_quote">On Jun 16, 2011 6:03 PM, "Michael Trausch" <<a href="mailto:mike@trausch.us">mike@trausch.us</a>> wrote:<br type="attribution">> It is possible that nondeterminism depends on the SATA chipset, but I have a<br>
> system in production where the drives are enumerated in the order they<br>> appear to the kernel. Furthermore... if I pull sdb out (yay hot swap) and<br>> put it back in, it'll get sdg (sdf is the last one assigned at boot time in<br>
> this system).<br>> <br>> I don't really worry about it, though, because I use the serial numbers, not<br>> the node names.<br>> <br>> --<br>> Sent from my phone... a G2 running CM7 nightlies!<br>
> On Jun 16, 2011 4:56 PM, "David Tomaschik" <<a href="mailto:david@systemoverlord.com">david@systemoverlord.com</a>> wrote:<br>>> On Thu, Jun 16, 2011 at 4:38 PM, Jim Kinney <<a href="mailto:jim.kinney@gmail.com">jim.kinney@gmail.com</a>> wrote:<br>
>>> Sort of but not quite all correct. The hard drive chain was not changed<br>>>> totally. Non-deterministic for pci-bus devices but still traceable with<br>>>> lspci. Always non-d for usb (pita)<br>
>>><br>>>> Hard drives follow a specific pattern:<br>>>><br>>>> BIOS spills data to sys about drive locations. Bus num followed by device<br>>>> num. That doesn't change 'cause it can't.<br>
>>><br>>>> If a new drive is inserted at a lower bus num than other drives, it gets<br>>>> called sda. Move the drive in socket 0 to socket 5 and it now is called<br>> sdf.<br>>>><br>
>>> But so what?! Most distros use UUID anyway so you can move your drives<br>>>> around between boots and it'll still work as long as the BIOS knows where<br>>>> the /boot drive is. Cool thing is error messages will reflect the current<br>
>>> configuration.<br>>>><br>>>> So if the bad drive is moved from sdb to sdf, on reboot the error will<br>>>> reflect the bad drive is sdf.<br>>>><br>>>> So as long as drives stay plugged in the same, detection will be<br>
>>> deterministic but the name is not. Remember, empty sockets 0-5 and a<br>> single<br>>>> drive in 5 will be called sda.<br>>>><br>>>> Thus by looking for the next to lowest numbered drive will reveal sdb,<br>
> the<br>>>> failed drive in the OP. :-)<br>>><br>>> Unless, of course, there are udev rules that specify otherwise.<br>>> Serial # is still the most reliable way to be CERTAIN of what you're<br>
>> pulling. Removing the wrong drive from a RAID can make your day very<br>>> bad. (I've placed SN labels on the visible end of drives in my home<br>>> system for exactly this reason. Or paranoia. Or because I like<br>
>> labels. Take your pick.)<br>>><br>>><br>>><br>>> --<br>>> David Tomaschik, RHCE, LPIC-1<br>>> System Administrator/Open Source Advocate<br>>> OpenPGP: 0x5DEA789B<br>>> <a href="http://systemoverlord.com">http://systemoverlord.com</a><br>
>> <a href="mailto:david@systemoverlord.com">david@systemoverlord.com</a><br>>><br>>> _______________________________________________<br>>> Ale mailing list<br>>> <a href="mailto:Ale@ale.org">Ale@ale.org</a><br>
>> <a href="http://mail.ale.org/mailman/listinfo/ale">http://mail.ale.org/mailman/listinfo/ale</a><br>>> See JOBS, ANNOUNCE and SCHOOLS lists at<br>>> <a href="http://mail.ale.org/mailman/listinfo">http://mail.ale.org/mailman/listinfo</a><br>
</div>