[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

[linux-lvm] [Bug#212483: vgscan fails at least with raids inside raids]



-- 

patrick

--- Begin Message ---
Package: lvm10
Version: 1:1.0.7-7
Severity: important

vgscan tries to scan every device and by doing so it fails at least under
some circumstances.

The circumstances on which I have found this are these: I have a raid0
device named md0 which is itself part of a raid5 device named md2.

This is what happens: I create a pv on the raid5 device and then define a
group named raid5vg inside and a couple of logical volumes inside of this, I
format the lvs and I even mount them and play with the system, everything is
fine till I boot up again, I see that starting the lvm fails because of
vgscan failing.

What happens to vgscan is that it is finding stuff about raid5vg in both md0
and md2, this is caused by md0 being part of md2, I have tried changing the
position of md0 in the raid (making it the last of the 3 devices I have in
the raid5, instead if the first one) but same thing happened.

To prove that my supposition is correct I have done this:

I run vgscan with all the raids started and i get:

vgscan -- reading all physical volumes (this may take a while...)
vgscan -- ERROR "lv_read_all_lv(): number of LV" can't get data of volume group
"raid5vg" from physical volume(s)
vgscan -- "/etc/lvmtab" and "/etc/lvmtab.d" successfully created
vgscan -- WARNING: This program does not do a VGDA backup of your volume group

Then I stop the raid0 by: stopping the raid5, then the raid0 and then
starting the raid5 in a dregradated mode. These are the commands I issue:

raidstop /dev/md2
raidstop /dev/md0
raidstart /dev/md2

And run again vgscan, but now md0 doesn't exist as it is not started, so I
get...

vgscan -- reading all physical volumes (this may take a while...)
raid5: switching cache buffer size, 4096 --> 1024
vgscan -- found inactive volume group "raid5vg"
vgscan -- "/etc/lvmtab" and "/etc/lvmtab.d" successfully created
vgscan -- WARNING: This program does not do a VGDA backup of your volume group

If I start md0 again, I get the error again.

I compiled lvm10 in debug mode and run vgscan -v --debug in the case where
it fails, I'm attaching the output to this report.

I think that vgscan shouldn't have any kind of problem in this situation, I
mean, maybe he cannot get the info about raid5vg from md0, but he surely can
get it from md2.

Regards...

-- System Information:
Debian Release: testing/unstable
Architecture: i386
Kernel: Linux man 2.4.22 #1 Mon Sep 22 00:43:52 CEST 2003 i586
Locale: LANG=C, LC_CTYPE=C

Versions of packages lvm10 depends on:
ii  debconf                       1.3.14     Debian configuration management sy
ii  file                          4.03-3     Determines file type using "magic"
ii  libc6                         2.3.2-8    GNU C Library: Shared libraries an
ii  lvm-common                    1.5.9.3    The Logical Volume Manager for Lin

-- debconf information:
* lvm10/warning: 

Attachment: /tmp/vgscan.debug.gz
Description: Binary data


--- End Message ---

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]