Does not work on current coreos stable

[natami]

Following the instructions on a clean newly provisioned coreos installation:

sudo su -
git clone [email protected]:packethost/packet-block-storage.git
cp ./packet-block-storage/packet-block-storage-* /usr/bin/
chmod u+x /usr/bin/packet-block-storage-*

• although the /usr/bin/ is a readonly folder, i placed the scripts under home/core.

When executing, i get the following error:
portal: 10.144.136.91 iqn: iqn.2013-05.com.daterainc:tc:01:sn:bf852caxxxxxxxxxxx
Oct 26 19:37:17 | volume-a686xxxx: ignoring map
Oct 26 19:37:17 | volume-a686xxxx: ignoring map
portal: 10.144.140.241 iqn: iqn.2013-05.com.daterainc:tc:01:sn:bf852caxxxxxxxxxxx
Oct 26 19:37:20 | volume-a686xxxx: ignoring map
Oct 26 19:37:20 | volume-a686xxxx: ignoring map
Error: Block device /dev/mapper/volume-a686xxxx is NOT available for use

[Promaethius]

Having the same issue on coreos-stable. The disclaimer for their script does omit CoreOS.

After enabling the verbose version, however, the volume DOES actually turn up not as its volume name but as the WWID rather than the device alias under /dev/mapper. In other words, the device is mounting but not with the name you'd expect.

After researching the topic, I'm happy to say that the entire script is fixable with a single line of code inserted after the configuration of multipath (or for me around line 272:

multipathd reconfigure

This forces multipath to reload the configuration which wasn't picked up immediately after the service was enabled and started and allows for the user_friendly_names option to kick in. This binds the correct volume name to the WWID which shows up under /dev/mapper.

Hope this helps future users of CoreOS.

[dlaube]

@Promaethius Thank you for digging into this for CoreOS! The change looks pretty good but seems to produce the following error on CentOS 7/Ubuntu 14;

reload: volume-XXXXXXX (36001405e408064ea9ab41749e79ef33c) undef DATERA ,IBLOCK
size=300G features='0' hwhandler='1 alua' wp=undef
-+- policy='round-robin 0' prio=1 status=undef |- 12:0:0:0 sdc 8:32 active ready running - 13:0:0:0 sdd 8:48 undef ready running
ux_socket_connect: Connection refused
Block device /dev/mapper/volume-XXXXXXX is available for use

We're happy to accept a PR if you can make the call to multipathd reconfigure dependent upon the distro --in this case, just for CoreOS. Perhaps parse /etc/lsb-release and/or make this conditional on the existence of DISTRIB_ID=CoreOS ?

[nathangoulding]

@dlaube can you take a look and test the updated change?

[alanpearce]

multipathd reconfigure also appears to be required in debian 9.

[Promaethius]

@alanpearce thanks I'll add it to my pr