Unable to install Ubuntu 16.04 on Tuleta with multipath setup

Bug #1546606 reported by bugproxy
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
hw-detect (Ubuntu)
Fix Released
High
Mathieu Trudel-Lapierre

Bug Description

---Problem Description---
Unable to succeed in installing Ubuntu 16.04 on Tuleta with multipath setup; installation hung at scanning disks

== Comment: #11 - Kevin W. Rudd - 2016-02-05 13:13:11 ==

It looks like the issue in this case involves an RDAC based storage controller and I/O errors on the ghost paths.

== Comment: #23 - Mauricio Faria De Oliveira - 2016-02-17 07:48:23 ==
Hi Canonical,

This patch resolves this problem. It's been tested on 16.04.
Please consider it for uploading.

@taco-screen-team, I believe @mathieu-tl would be assigned to this one, it it helps.

Testing udeb package:

 ~ # wget http://ausgsa.ibm.com/~mauricfo/public/bugs/bz136625/v1/disk-detect_1.114ubuntu1scsidh1_all.udeb
 ~ # udpkg -i disk-detect_1.114ubuntu1scsidh1_all.udeb

 Detect disks happened.

It is working correctly.

1) The SCSI DH modules are correctly loaded and attached (see references count) to SCSI disk.

 ~ # lsmod | grep scsi
 scsi_transport_fc 71759 1 lpfc
 scsi_dh_alua 9962 16
 scsi_dh_hp_sw 5252 0
 scsi_dh_emc 9486 0
 scsi_dh_rdac 9656 32

2) The 'Scanning disks' dialog show I/O error dialogs rather than a silent, extremely long delay.

 [!] Partition disks

 Error fsyncing/closing /dev/sdag: Input/output error

 Warning!

 Retry
 Ignore

 <Go Back>

== Comment: #26 - Mauricio Faria De Oliveira <email address hidden> - 2016-02-17 09:20:28 ==
Quick summary.

The long delays are caused due to dmraid scanning the RDAC unowned paths to the storage unit (seen in multipath -l as 'ghost' paths).

They happen in 2 points: during hw-detect/disk-detect and partman/init.d/30parted (Scanning disks).

If you have multipath enabled in the installation,
the 2nd delay doesn't happen as the individual SCSI disks (/dev/sdX) are not considered by partman,
but the 1st delay does occur *anyway*, as /bin/disk-detect does call dmraid to scan available devices.

So, the fix is beneficial for both scenarios (multipath enabled and disabled; e.g., user forgot to/didn't know he should enable it).

Revision history for this message
bugproxy (bugproxy) wrote : Ubuntu boot/install log

Default Comment by Bridge

tags: added: architecture-ppc64le bugnameltc-136625 severity-high targetmilestone-inin1604
Revision history for this message
bugproxy (bugproxy) wrote : Patch for hw-detect to load SCSI device handler modules

Default Comment by Bridge

Changed in ubuntu:
assignee: nobody → Taco Screen team (taco-screen-team)
Kevin W. Rudd (kevinr)
affects: ubuntu → debian-installer (Ubuntu)
Revision history for this message
Mathieu Trudel-Lapierre (cyphermox) wrote :

Reassigning to hw-detect -- there's a patch provided already, and the logic seems sound. I'm reviewing the bug and the included patch to upload the fix.

affects: debian-installer (Ubuntu) → hw-detect (Ubuntu)
Changed in hw-detect (Ubuntu):
assignee: Taco Screen team (taco-screen-team) → Mathieu Trudel-Lapierre (mathieu-tl)
status: New → Triaged
importance: Undecided → High
Changed in hw-detect (Ubuntu):
status: Triaged → In Progress
Revision history for this message
bugproxy (bugproxy) wrote : Patch for multipath-tools to load SCSI device handlers before SCSI low-level device drivers (boot time)

------- Comment on attachment From <email address hidden> 2016-02-18 13:11 EDT-------

Hi @mathieu-tl,

I'm attaching a patch for multipath-tools as well,
so to avoid any delays and more SCSI errors during boot time.

Tested on another system, attached via FC to DS4800 storage (which uses RDAC), and with IPR disks (which uses ALUA).

Thanks

Revision history for this message
bugproxy (bugproxy) wrote : Comment bridged from LTC Bugzilla

------- Comment From <email address hidden> 2016-02-18 13:25 EDT-------
@mathieu-tl,

(In reply to comment #32)
> I'm attaching a patch for multipath-tools as well,
> so to avoid any delays and more SCSI errors during boot time.
>
> Tested on another system, attached via FC to DS4800 storage (which uses
> RDAC), and with IPR disks (which uses ALUA).

Here's the boot time verification:

<...>
Loading, please wait...
Begin: Loading multipath modules ... [ 2.256763] device-mapper: multipath: version 1.10.0 loaded
Success: loaded module dm-multipath.
done.
Begin: Loading multipath hardware handlers ... [ 2.259401] alua: device handler registered
Success: loaded module scsi_dh_alua.
[ 2.261424] rdac: device handler registered
Success: loaded module scsi_dh_rdac.
done.
starting version 229
<...>

$ lsmod | grep scsi
scsi_transport_fc 71759 1 lpfc
scsi_dh_rdac 9656 32
scsi_dh_alua 9962 16

------- Comment From <email address hidden> 2016-02-18 13:29 EDT-------
*** Bug 125894 has been marked as a duplicate of this bug. ***

Revision history for this message
Launchpad Janitor (janitor) wrote :

This bug was fixed in the package hw-detect - 1.114ubuntu2

---------------
hw-detect (1.114ubuntu2) xenial; urgency=medium

  * disk-detect.sh: load SCSI device handlers before SCSI low-level
    device drivers (LP: #1546606).

 -- Mauricio Faria de Oliveira <email address hidden> Wed, 17 Feb 2016 10:19:45 -0200

Changed in hw-detect (Ubuntu):
status: In Progress → Fix Released
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.