Bug 621070

Created attachment 436460 [details]
output.log of storage test on RHEL6 ppc64

This means the installation was not configured properly as the drive  was already in use by LVM and thus couldn't be used for the destructive testing.

https://fedorahosted.org/v7/wiki/HwCertHowToTestProcedures#storage

But the installation was finished automatically through the Beaker. And the storage test on RHEL5.5 on the same machine is successful. So the failure on RHEL6 is really due to the installation was not configured properly?

Technically yes, but there is also a bug and probably a design change to consider.

Swap is used in RHEL5 because swap must be a physical block device and the storage test needs to write to the physical block device, since an lvm can be scattered across drives you could end up never testing the drive in question.

In RHEL6 swap can be on lvm which is where the problem comes in.  In RHEL6 swap is on lvm by default; when the storage test runs it decides swap is it's best bet to access a physical drive so it tries only to figure out later that it's on lvm and that it can't be used.

The bug is that it leaves this swap partition offline resulting in a system now being configured to fail the memory test.  The solution is the memory test should 1st check to verify the swap partition it would like to use is not part of an lvm set.

The partition layout is in the user control so this could all be avoided as it was with RHEL5 by a creating a conforming layout that provided either blank space or a swap partition on the physical drive block device.

Obviously this creates a situation where we probably want to think about a design change for the storage test because it will be painful if the default behavior and valid configuration of RHEL6 is not certifiable.  Depending on the decisions of that discussion we may need to provide vendors with a work around or introduce some new code to address.

The reason swap doesn't get restored is that when the test calls mkswap (line 210, storage.py), mkswap has output on standard error:

# mkswap /dev/dm-1
mkswap: /dev/dm-1: warning: don't erase bootbits sectors
        on whole disk. Use -f to force.
Setting up swapspace version 1, size = 10371068 KiB
no label, UUID=3e0210c4-6340-44d0-b081-8e2e783660f4

mkswap seems to succeed.  The problem is that a V7CommandException is raised due to output on stderr, so the next call to swapon (line 211) does not happen.

This warning seems spurious to me.   I'd rather we didn't pass in -f, so instead we could print stderr as a warning for this command.   Also, exception handling needs to change so that swapon is called in all situations.

This bug is independant of arch - changing summary accordingly.

Created attachment 439777 [details]
storage test patch changing failures on mkswap to warnings

Re-run storage test on server s390x with RHEL6.0-20100822.n.0 and R20. It still failed with error "dasdd is currently in use by LVM.It cannot be tested. You may need to reinstall." The log shows that "file descriptor leaked on lvs invocation". It did not appear before.

Created attachment 440625 [details]
storage test log on RHEL6 s390x with R20

(In reply to comment #11)
> Re-run storage test on server s390x with RHEL6.0-20100822.n.0 and R20. It still
> failed with error "dasdd is currently in use by LVM.It cannot be tested. You
> may need to reinstall." The log shows that "file descriptor leaked on lvs
> invocation". It did not appear before.

The "in use ny LVM" error is a test set-up/configuration issue.  This bug is where the storage test tries to use an lvm swap partition, then fails, and also fails to restore swap, causing other failures in the memory and core test.

I'm moving this back to ON_QA - please verify the lvm/swap issue is fixed.

I've tested in i386/x86_64, ppc64 with r20 and r21, lvm swap partition can be used for testing and also can be restored. 

Seem that the swap partition can't be enabled nor reinitialize via `mkswap' in Beaker so that the problem in #comment 11 appears. This should be a different problem. Talked to qcui and we decide to move this bug to VERIFIED

An advisory has been issued which should help the problem
described in this bug report. This report is therefore being
closed with a resolution of ERRATA. For more information
on therefore solution and/or where to find the updated files,
please follow the link below. You may reopen this bug report
if the solution does not work for you.

http://rhn.redhat.com/errata/RHBA-2010-0702.html

    Technical note added. If any revisions are required, please edit the "Technical Notes" field
    accordingly. All revisions will be proofread by the Engineering Content Services team.
    
    New Contents:
The V7-1.2-14 storage test no longer fails on IBM System z Red Hat Enterprise Linux 6 and 64-bit PowerPC Red Hat Enterprise Linux 6.