On Wednesday 4 January 2012 at 19:23, Ritesh Raj Sarraf wrote:
> On 01/04/2012 11:39 PM, Laurent Bigonville wrote:
> > Instead of loading all kernel modules, it could maybe be interesting
> > to add a config file in /etc/initramfs-tools/conf.d/ to let the user
> > select which hardware handler he wants to load before udev?
> >
>
>
> Yup. That'd be the best in this case.
>
> Looks like we already are doing something like that:
>
> ====8<================8<====================8<===
>
> maybe_break pre-multipath
> VERBOSITY=0
> MP_MODULES="dm-multipath dm-emc"
>
> if [ ! -e /sbin/multipath ]; then
> exit 0
> fi
>
> verbose && log_begin_msg "Loading multipath modules"
> for module in ${MP_MODULES}; do
> if modprobe --syslog "$module"; then
> verbose && log_success_msg "loaded module ${module}."
> else
> log_failure_msg "failed to load module ${module}."
> fi
> done
> verbose && log_end_msg
>
> ====8<================8<====================8<===
>
>
> If anyone can try this out, it'll be great.
>
>
> --
> Ritesh Raj Sarraf | http://people.debian.org/~rrs
> Debian - The Universal Operating System
>
>
Hi,
I would also like to add that in the first place we are not interested in SAN
boot. The multipath script with MP_MODULES in local-top is included with
multipath-tools-boot, so that wouldn't be a solution in our case as it doesn't
make sense to install the multipath-tools-boot package in this situation since
I don't really need multipath configuration in initramfs. I only need it after
the system has booted from its local disks.
Since MODULES=most in /etc/initramfs-tools/initramfs.conf this means that our
HBA driver (qla2xxx) is already included in initramfs. Because multipath and
scsi_dh_rdac modules are not loaded at this point, the scsi errors cause delays.
Currently, I worked around this by putting MODULES=deb and load the
scsi_dh_rdac driver in /etc/initramfs-tools/modules, so that the scsi_dh_rdac
driver is certainly loaded before qla2xxx. Of course this wouldn't work in a
boot-from-san situation. But configuring it like this works around the whole
problem.
I agree with Laurent that loading the hardware handler in local-top is too late
(I tested this), and should be done in init-top to make it work.
Best regards,
--
Frido Roose
Sent with Sparrow (http://www.sparrowmailapp.com/?sig)
