Bug#580972: multipath-tools: unnecessary call to multipath in /etc/udev/rules.d/z60_multipath.rules

[Ferenc Wagner]

Apollon Oikonomopoulos <apoikos at gmail.com> writes:

> On 18:08 Mon 10 May, Guido Günther wrote:
>
>> This is also necessary for the initramfs since we don't run multipathd
>> there. Testing if multipathd is runnning looks like a good option
>> though. It'd be even nicer if we could only run that rule if the former
>> one fails.
>
> Actually, since there is already an initramfs-tools hook, how about adding a
> rule only specific to the initramfs, omitting the socket rule altogether, and
> then keeping only the socket rule in the system's /etc/udev/rules.d/? IIRC,
> udev is restarted during S, so I suppose it loads all rules from scratch.

Well, I think it would make more sense to run multipathd in the
initramfs instead.  System boot time is already suffering enough, with 5
LUNs accessible on 4 paths each I get messages like

[  340.353997] sd 0:0:0:3: emc: at SP A Port 3 (owned, default SP A)

during the initramfs phase, which is almost 6 minutes of uptime spent on
discovering and scanning devices!  Hmm, this seems really excessive,
maybe the slow serial console is more responsible of it, hand in hand
with the stupid unavoidable kernel partition scanning spewing out
thousands of verbose device error messages...
-- 
Regards,
Feri.