Package Details: multipath-tools 0.7.7-1

Git Clone URL: (read-only)
Package Base: multipath-tools
Description: Multipath tools for Linux (including kpartx)
Upstream URL:
Licenses: GPL2
Conflicts: multipath-tools-git
Submitter: xyproto
Maintainer: Bevan
Last Packager: Bevan
Votes: 84
Popularity: 2.877920
First Submitted: 2013-01-24 14:33
Last Updated: 2018-05-11 17:04

Dependencies (7)

Sources (1)

Latest Comments

ozz commented on 2018-06-21 15:48

It works building with yaourt and I get the "cloning in to..." message like it's suppose to instead of just erroring out without doing anything.

I believe it's some sort of environmental issue (temp directories and such). Possibly a bug in makepkg itself.

ozz commented on 2018-06-20 20:27

Yes, base-devel installed. "pacman -S --needed base-devel" skips all of them. Yes, cd to the directory before makepkg (btw, makepkg won't do anything if that isn't done :)

Bevan commented on 2018-06-20 20:23

Since it's a fresh installation: Are all packages from the base-devel group installed?

pacman -S --needed base-devel

And to be sure: You run git clone, then cd into multipath-tools and then run makepkg, right?

ozz commented on 2018-06-20 20:03

That directory did not exist until running "makepkg". Literally all I do is "git clone" then "makepkg" and it fails with the error I posted below. Removing the directory doesn't change anything. I wonder if the source=("multipath-tools::git+http... is the problem. Maybe I'm missing a handler that should be checked for. This is a clean new fresh Arch install. Whatever the case, something is broken.

Bevan commented on 2018-06-20 19:48

ozz: Seems like in the directory where you are trying to build the package, there is already a folder called "multipath-tools". That's why makepkg cannot checkout the source git at the same location. Solution is to remove that subdirectory.

ozz commented on 2018-06-20 14:46

Doesn't build out of the box.

$ makepkg ==> Making package: multipath-tools 0.7.7-1 (Wed 20 Jun 2018 10:44:35 AM EDT) ==> Checking runtime dependencies... ==> Checking buildtime dependencies... ==> Retrieving sources... ==> ERROR: multipath-tools/multipath-tools is not a clone of Aborting...

Bevan commented on 2018-01-06 18:08

I think you can also add specific devices to your blacklist exception. The fact that LVM uses the devices before multipath can may suggest that adding some directives to the systemd service so that lvm only starts after multipathd may also help.

But to be honest, I have only very limited own experience with multipath setups. Maybe you should ask on the dm-devel mailing list about this.

jab416171 commented on 2018-01-06 17:06

I tried adding that, and now this is my blacklist_exceptions (output from multipath -t):

blacklist_exceptions { property "(SCSI_IDENT_.*|ID_WWN)" property "(SCSI_IDENT_|ID_WWN)" }

After rebooting, I only get 2 multipath devices that show up. There were 6 before I rebooted. The other LUNs show up, but just the individual partitions, not the "alias" I defined in multipath.conf.

Actually upon closer inspection, it appears lvm is finding and opening the devices before multipath can.

Bevan commented on 2018-01-06 10:04

jab416171: Have a look at the section "blacklist_exceptions section" in man multipath.conf. I guess your devices are missing the SCSI_IDENT_.* property. However, this is required since 0.6.0 already so I'm not sure why you didn't have the issue with 0.6.4:;a=commitdiff;h=d041591e42b69e2ff99d9cc5c1111c83ccde3207

RedHat disables this blacklisting in their distributions:

0004-RH-Remove-the-property-blacklist-exception-builtin.patch: Allow multipath to be used on devices without multiple paths. NAK'ed upstream, but requested by Red Hat

Since you can easilly change this behavior via blacklist_exceptions in multipath.conf, I wouln't like to deviate from upstream here though.

jab416171 commented on 2018-01-06 06:59

I just updated from 0.6.4 to 0.7.4, and now only two of my LUNs are showing up. There should be about 16. multipath -v3 -ll mentions this: dm-x: blacklisted, udev property missing (once for each LUN). Two of the LUNs do work, interestingly, even though they are also 'blacklisted'.

All comments