Bug 155550

Summary:	multipath -l rescans all paths
Product:	[Fedora] Fedora	Reporter:	Lars Marowsky-Bree <lmb>
Component:	device-mapper-multipath	Assignee:	Alasdair Kergon <agk>
Status:	CLOSED NEXTRELEASE	QA Contact:
Severity:	medium	Docs Contact:
Priority:	medium
Version:	rawhide	CC:	agk, christophe.varoqui, dmo, egoggin, lmb, tranlan
Target Milestone:	---	Keywords:	FutureFeature
Target Release:	---
Hardware:	All
OS:	Linux
Whiteboard:
Fixed In Version:		Doc Type:	Enhancement
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:	2005-05-05 21:37:47 UTC	Type:	---
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:

Description Lars Marowsky-Bree 2005-04-21 12:09:37 UTC

multipath -l rescans all physical paths; which, if a path is failing/timed out
can take quite long and gets stuck easily.

I think it would be better if it just scanned the DM multipath tables and
reported what it finds there, w/o accessing the backing devices. This would also
be much, much faster.

Comment 1 Christophe Varoqui 2005-04-22 22:34:45 UTC

I could make a "light list" option, but not rescaning all paths will trim the
output from :
- prio (not printed now actually)
- path state (according to checkers)
- h/b/t/l
- devname (/dev/sda)

Now, path_discovery() have this new flag to fetch just what is necessary.
I personnaly think the bare minimum info to output in addition to DM table and
status are h/b/t/l and devname.

I could make that the output of '-l', '-ll' adding checkers, '-lll' adding prio

What do you think ?

Comment 2 Lars Marowsky-Bree 2005-04-25 12:17:45 UTC

Well, scanning the devname doesn't require to actually _open_ the device nodes
though (and thus doesn't cause IO delays); merely stat()ing them is enough, or
even scanning /proc/partitions for the name associated with the major:minor.

h/b/t/l likewise only requires to scan sysfs, w/o needing to access the physical
device.

And, when I'm asking for the current status, I want the kernel's view, not what
the path checker thinks ;-) (The path checker could be displayed as additional
information.)

So I think your proposal makes perfect sense.

Comment 3 Lars Marowsky-Bree 2005-04-25 12:18:26 UTC

Oh, actually: The path checker & prio status could be requested from the daemon
(if running), no?

Comment 4 Christophe Varoqui 2005-05-04 21:12:16 UTC

'-l', '-ll', as described above is merged in commit
b30da450994dc52b3db45b5268a6fb418094796b at korg.

I suggest closing this one, as prio values are only relevant as a debug info and
is available via '-v3+'. I think '-lll' is not required after all.

Comment 5 Alasdair Kergon 2005-05-05 21:41:31 UTC

Closed: using 'NEXTRELEASE' to indicate that it's resolved in multipath-tools
and will get included in Fedora next time we import your tarball.

It's still possible to continue the discussion here after it's closed, and it
can easily be re-opened if we do decide we want more changes.