Bug 169403

I guess this is fixed in U2/U6 already. The initsripts maintainer suggested to
look at this bug 148764.

Mike, this bug is being reported against RHEL 3.0 U6 which BZ #148764 states as 
having been fixed.

I believe the issue may be related to dealing with an emcpower pseudo device at 
this time. fsck() is trying to run before the emcpower pseudo can be created. 
And in there may lay the rub.

We've re-run the test against RHEL 3.0 U6 without the use of the emcpower 
psuedo device with the same results, fsck() on the native iSCSI device fails. 
Here are the server's particulars:

[root@libg184 root]# rpm -qa |grep init
initscripts-7.31.27.EL-1
iscsi-initiator-utils-3.6.3-3

[root@libg184 root]# uname -a
Linux libg184 2.4.21-37.ELsmp #1 SMP Wed Sep 7 13:28:55 EDT 2005 i686 i686
i386 GNU/Linux

The /dev/sd device is set in /etc/fstab with the fsck on boot flag and the 
_netdev option.

Just changing component to initscripts becuase there is not a iscsi component
anymore, and this does not look like a iscsi-initiator-utils or kernel problem.

I think the problem is that we have been running with fastboot. When this is set
rc.sysint

The fixes for RHEL3 and RHEL4 to the netfs scripts that were made in this BZ
148764 work and are ok. For RHEL4, and probably RHEL3, the problem seems to be
in rc.sysinit. I think some of use have been testing with fastboot set. With
this set, rc.sysinit does not fsck the /etc/fstab devices, and everything works
fine. When fastboot is not set, the rc.sysinit fsck code blows up on NFS, iscsi,
and other weirdo devices/filesystems. 

What is the proper fix here? Should rc.sysinit skip _netdev devices, or is there
a fstab variable we should be setting, or are we supposed to be running with
fastboot set? Thanks

_netdev devices are explicitly fscked in the netfs initscript, as I recall.
Don't set the auto-fsck flag in fstab, and things should be fine.

netfs does this

fsck -A -T -R -a -t opts=_netdev

But fsck still obeys the fs_passno numbering so if it is zero then fsck (run
from rc.sysinit and netfs) will not check the fs.

Do you think this is a fsck bug. Is the "-t opts=_netdev" argument supposed to
mean that fsck is supposed to check all entries with _netdev ignoring the fs_passno.

Oops,  I misunderstood the behavior.

Does adding -t opts=!_netdev in rc.sysinit fix it?

yeah that works.

Will this fix go into RHEL 3.0 U7? Is a similar fix needed for RHEL 4.0?

We cloned this bug for RHEL4
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=172804
It looks like you did not get added becuase the reported does not get copied
over when someone else clones it.

And it looks like Bill added the fixes to the newest builds so I think it will
be in both RHEL3 U7 and RHEL4 U3.

Huh, not sure what the problem was before, but I did a clean install and updated
the initscripts and it works. Sorry about that.

Clean install for RHEL 3.0 U7 or RHEL 4.0 U3?

Hey Wayne, it was for RHEL3 but you can disregard my comment. It was in regards
to some other chitchat on the bug.

Created attachment 125416 [details]
boot.log

boot.log from server

Created attachment 125417 [details]
messages from server

What should be the expected results when the changes take affect? The 
filesystems mounts; however, there are errors reported by fsck() and it doesn't 
appear that fsck is run on the filesystem after it does finally mount. See the 
preceeding attachments for boot.log and messages.

Regards,
Wayne.

Server:
[root@l82bi121 root]# cat /etc/redhat-release
Red Hat Enterprise Linux AS release 3 (Taroon Update 7 Beta)
[root@l82bi121 root]# uname -a
Linux l82bi121.lss.emc.com 2.4.21-40.EL #1 SMP Thu Feb 2 22:14:00 EST 2006 
x86_64 x86_64 x86_64 GNU/Linux
[root@l82bi121 root]# rpm -qa | grep init
initscripts-7.31.30.EL-1
xinitrc-3.32-1
SysVinit-2.85-4.4
mkinitrd-3.5.13.6-1
iscsi-initiator-utils-3.6.3-3
[root@l82bi121 root]# cat /etc/fstab
LABEL=/                 /                       ext3    defaults        1 1
LABEL=/boot             /boot                   ext3    defaults        1 2
none                    /dev/pts                devpts  gid=5,mode=620  0 0
none                    /proc                   proc    defaults        0 0
none                    /dev/shm                tmpfs   defaults        0 0
/dev/sda3               swap                    swap    defaults        0 0
/dev/cdrom              /mnt/cdrom              udf,iso9660 
noauto,owner,kudzu,ro 0 0
/dev/fd0                /mnt/floppy             auto    noauto,owner,kudzu 0 0
128.221.37.21:/pub      /pub    nfs     defaults 0 0
128.221.37.21:/pub/wayne/extra  /extra  nfs     defaults 0 0
172.23.189.135:/wide_open/misc/open     /extra/open     nfs     defaults 0 0
/dev/emcpowera1         /zoner/emcpowera        auto    _netdev         1 1
[root@l82bi121 root]# df
Filesystem           1K-blocks      Used Available Use% Mounted on
/dev/sda2             68445304   4147292  60821148   7% /
/dev/sda1               101089     13926     81944  15% /boot
none                   1977152         0   1977152   0% /dev/shm
128.221.37.21:/pub   138733088 128333024  10400064  93% /pub
128.221.37.21:/pub/wayne/extra
                     138733088 128333024  10400064  93% /extra
172.23.189.135:/wide_open/misc/open
                     413815392  72922760 319871984  19% /extra/open
/dev/emcpowera1        8620700      2468   8180320   1% /zoner/emcpowera
[root@l82bi121 root]#

Is there fsck error we are talking about the one where it cannot autodetect the
/zoner/emcpowera partition? What happens when you just put the filesystem type
in fstab?

[root@l82bi121 root]# fsck /dev/emcpowera1
fsck 1.32 (09-Nov-2002)
e2fsck 1.32 (09-Nov-2002)
/dev/emcpowera1 is mounted.

WARNING!!!  Running e2fsck on a mounted filesystem may cause
SEVERE filesystem damage.

Do you really want to continue (y/n)? no

check aborted.
[root@l82bi121 root]# umount /zoner/*
umount: /zoner/emcpowerb: not mounted
umount: /zoner/emcpowerc: not mounted
umount: /zoner/emcpowerd: not mounted
umount: /zoner/emcpowere: not mounted
umount: /zoner/emcpowerf: not mounted
umount: /zoner/emcpowerg: not mounted
umount: /zoner/emcpowerh: not mounted
umount: /zoner/emcpoweri: not mounted
umount: /zoner/emcpowerj: not mounted
umount: /zoner/emcpowerk: not mounted
umount: /zoner/emcpowerl: not mounted
umount: /zoner/emcpowerm: not mounted
umount: /zoner/emcpowern: not mounted
umount: /zoner/emcpowero: not mounted
umount: /zoner/emcpowerp: not mounted
[root@l82bi121 root]# df
Filesystem           1K-blocks      Used Available Use% Mounted on
/dev/sda2             68445304   4153792  60814648   7% /
/dev/sda1               101089     13926     81944  15% /boot
none                   1977152         0   1977152   0% /dev/shm
128.221.37.21:/pub   138733088 125748160  12984928  91% /pub
128.221.37.21:/pub/wayne/extra
                     138733088 125748160  12984928  91% /extra
172.23.189.135:/wide_open/misc/open
                     413815392  74102784 318691960  19% /extra/open
[root@l82bi121 root]# fsck /dev/emcpowera1
fsck 1.32 (09-Nov-2002)
e2fsck 1.32 (09-Nov-2002)
/dev/emcpowera1: clean, 169/1095584 files, 35006/2189564 blocks
[root@l82bi121 root]#

It doesn't seem to be an fsck() issue. I'll change the fstype in /etc/fstab to 
see if that mitigates the failure.

Hi Mike.

With the following entry in /etc/fstab I see an error message from 
init:mount:fsck: "Could not determine filesystem type for /dev/emcpowera1" and 
init:netfs:fsck is successful.

/dev/emcpowera1         /zoner/emcpowera        auto    _netdev         1 1

If I change the entry in /etc/fstab to the following there are no error 
messages from init:netfs:fsck and init:netfs:fsck is successful.

/dev/emcpowera1         /zoner/emcpowera        ext2    _netdev         1 1

Regards,
Wayne.

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 the 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/RHSA-2006-0015.html