Re: multipathd ignoring dev_loss_tmo setting

[Benjamin Marzinski <bmarzins@redhat.com>]

On Fri, Mar 01, 2019 at 10:04:52AM +0000, Martins, Bruno O wrote:
> On Thu, 2019-02-28 at 13:53 -0600, Benjamin Marzinski wrote:
> > On Thu, Feb 28, 2019 at 11:38:22AM +0000, Martins, Bruno O wrote:
> > > Hello guys,
> > > 
> > > I am trying to modify /etc/multipath.conf on my system so that the
> > > parameter 'dev_loss_tmo' is changed from the default value.
> > > 
> > > However, when checking the value currently in use I am getting the
> > > wrong value (which is '30') for some of the remote ports:
> > > 
> > 
> > Are you sure those rports are used by multipath devices? multipath
> > only
> > changes dev_loss_tmo for rports associated with a multipath path
> > device.
> > 
> > -Ben
> >  
> 
> Hi Benjamin,
> 
> Thanks for your reply!
> 
> I believe they are:
> 
> [ 10:02:45 ] root@myhost:~# multipath -ll | grep 3:0:3
>   |- 3:0:3:12  sdblc 128:1568 active ready running
>   |- 3:0:3:13  sdbnu 132:1664 active ready running
>   |- 3:0:3:18  sdbre 66:1792  active ready running
>   |- 3:0:3:1   sdbkg 70:1728  active ready running
>   |- 3:0:3:2   sdbnv 132:1680 active ready running
>   |- 3:0:3:20  sdbrg 66:1824  active ready running
>   |- 3:0:3:17  sdbpg 134:1760 active ready running
>   |- 3:0:3:16  sdbpf 134:1744 active ready running
>   |- 3:0:3:11  sdbkf 70:1712  active ready running
>   |- 3:0:3:19  sdbrf 66:1808  active ready running
>   |- 3:0:3:14  sdbnw 132:1696 active ready running
>   |- 3:0:3:15  sdbpe 134:1728 active ready running
> 
> Is this the best way to check that information?

The scsi HBTL isn't guaranteed to line up with the rport id.  To find this
out you can either run

# ls -l /sys/block

and then check the rport for your path devices from the link
destination.

For instance

[root@ask-07 block]# ls -l /sys/block
<snip>
lrwxrwxrwx. 1 root root 0 Feb 21 02:45 sdb -> ../devices/pci0000:00/0000:00:0a.0/0000:06:00.0/host16/rport-16:0-0/target16:0:0/16:0:0:0/block/sdb
lrwxrwxrwx. 1 root root 0 Feb 21 02:45 sdc -> ../devices/pci0000:00/0000:00:0a.0/0000:06:00.0/host16/rport-16:0-0/target16:0:0/16:0:0:1/block/sdc
lrwxrwxrwx. 1 root root 0 Feb 21 02:45 sdd -> ../devices/pci0000:00/0000:00:0a.0/0000:06:00.1/host17/rport-17:0-0/target17:0:0/17:0:0:0/block/sdd
lrwxrwxrwx. 1 root root 0 Feb 21 02:45 sde -> ../devices/pci0000:00/0000:00:0a.0/0000:06:00.1/host17/rport-17:0-0/target17:0:0/17:0:0:1/block/sde
<snip>
 
Here, sdb and sdc are using rport-16:0-0, and sdd and sde are using
rport-17:0-0

Otherwise, you can pick an rport with the wrong dev_loss_tmo, and check

/sys/class/fc_remote_ports/<rport>/device/<target>/

In there there will be a number of scsi HBTL identifiers, for example

# ls /sys/class/fc_remote_ports/rport-16\:0-0/device/target16\:0\:0/
16:0:0:0  16:0:0:1  fc_transport  power  subsystem  uevent

If these HBTL ids (16:0:0:0 and 16:0:0:1) are for multipath path
devices, then multipath should be updating dev_loss_tmo for them. In my
example above, the scsi HBTL does correspond to the rport id, but this
isn't always the case.

-Ben

> BR,