multipath -p multibus gives errors like: /lib/udev/scsi_id: invalid option -- s The default getuid_callout in /etc/multipath.conf (i.e. the one used by multipath if getuid_callout is not explicitly given) uses the -s option, modifying it to: getuid_callout "/lib/udev/scsi_id -g -u -d /dev/%n" fixes the problem since scsi_id from >udev-124 no longer uses the -s option. See also http://www.google.com/search?hl=en&q=%2Flib%2Fudev%2Fscsi_id%3A%20invalid%20option%20--%20s&btnG=Google+Search . Perhaps an emerge time check and modification could be made to /etc/multipath.conf or ewarn could be added to the ebuild to notify users of this problem.
Fixed in upstream in d7329829 on 2008-08-19 20:54:25 +0200 and is probably included in next release.
I'd suggest splitting out a new revision 0.4.8 ebuild which simply removes the -s from the default and lock the udev version to >=124 via a blocker, so (newer) portage can automatically resolve it and ensure the right version is used.
I know that some people dislike this, but I'm setting the severity to critical, since it can completely stop multipath-enabled devices from being correctly setup, in the default configuration, no less. We don't want to be killing our SAN performance, now, do we?
> Fixed in upstream in d7329829 on 2008-08-19 20:54:25 +0200 and is probably > included in next release. Well, please tell me where this is fixed? I've used sources from here http://christophe.varoqui.free.fr/ and still the bug exists? Patch would be very helpful.
http://git.kernel.org/?p=linux/storage/multipath-tools/.git;a=summary and http://git.kernel.org/?p=linux/storage/multipath-tools/.git;a=commit;h=d7329829 upstream needs to stop slacking and cut a release. It's been 18+ months since the last one. If there isn't one by the end of the week, I'll cut it myself.
that breaks also xen vscsi
Just adding that changing getuid_callout to /lib/udev/scsi_id -g -u -d /dev/%n also works well for me.
Problem still exists 1+ year later?? The fix in original bug report works, please consider implementing. Thanks!
In 0.4.8-r1 now. Upstream needs to stop slacking and cut a release.
(In reply to comment #7) > Just adding that changing getuid_callout to /lib/udev/scsi_id -g -u -d /dev/%n > also works well for me. > If somebode else is lost, here is detailed hackish solution: http://forums.gentoo.org/viewtopic-t-738467-highlight-scsiid.html