Centos 5.1 bootup stops at Setting up Logical Volume Management

tiber · 02-07-2008, 12:01 PM

Hello everyone,

I'm at a loss here and hope someone can shed some light on this.

I have an Asus K8N-LR based system that I'm trying to install Centos 5.1 x86_64 on. I will be putting either Zimbra or Scalix on this box to serve as my personal email server.

I've experienced the following w/ two different drive setups - first with a SATA drive connected to the Mobo, and then with a PERC3 card w/ two scsi drives in RAID1. PERC card is set to Mass Storage in BIOS as necessary to work with this vintage kernel. I do not believe the drive setup to be the issue.

Both Scalix and Zimbra recommend disabling SELinux so the email server install and operation will go smoothly. I have tried multiple re-installs with different settings -
Fully Disabled - when I select this the system will not reboot... stops at Setting up LVM and sits there. Mouse will move.
Permissive - when I select this the same situation occurs.
Enabled - this will allow the system to reboot, but when I disable it in the GUI and then reboot I get the same problem.

I've tried this both with a default LVM system, and with manual partitioning not using LVM. LVM management still shows up, and still hangs there.

Waiting to disable until after a full yum update has completed does not correct the issue.

Am I missing something fundamental here? I've tried looking through this forum in depth but have not found anything.

EDIT -- If I do an interactive boot (press I) and select 'n' on lvm2, it will go through the rest of the boot process normally. So, I think the issue is narrowed down, but I'm at a loss as to how something seemingly basic can lock the system repeatedly.

Thanks in advance for any help/advice.

Eddie

Drakeo · 02-08-2008, 06:53 AM

yeap the problem is you have it all right but that lvm-PERC3 card, etc is not set to yes in your kernel. the kernel is the problem.the lvm must be built into the kernel. but you have to make sure they are enabled. On the newer kernels the hardware is loaded by the kernel it self. this speeds things up. so when you get to yum I hate it lol. it installs the stuff you need but it is not enabled in the kernel. I found this to be a big issue with the newer kernels but after building a few of them and enabling the stuff need I love it because boot time is cut in half. the old kernels loaded the system the system loaded the moduls through hot plug or udev. well the new kernels are huge weigh in about 4mb. if built right and it loads all your stuff during the boot of the kernel it self then hands it over to the system configuration. I use the slackware 2.6.21.5-smp-huge-smp and have compile also the 2.6.23.12-smp to be huge also. Dont know if this helped you but that is the problem. and also make sure you have CMOS (bios) set to none windows or non plug n play

Drakeo · 02-08-2008, 07:03 AM

oh yeah do not for get this is a scsi issue in the kernel settings also.

tiber · 02-08-2008, 10:10 AM

I appreciate the reply... any chance you or someone could elaborate a bit on how to go about that?

I am a novice w/ linux, and any detailed advice on how to go about this would be exceedingly helpful.
My master plan is to set up the mail server on this box, transfer my mail services over from the ClarkConnect 3.2 box I have, and then add my 3ware-based array out of the old linux server to this machine... then move /var to the 3ware array.

Thanks in advance for any help w/ details on this, or a pointer in the right direction.

Eddie

tiber · 02-08-2008, 10:59 AM

I now find that if I get to the gui before changing anything... go to the Security Level and Firewall applet... disable SELinux... the LVM applet will not load anything. If I go back and re-enable SELinux (enforcing), reload the LVM applet and the two volume groups magically show up.
This is frustrating to say the least... any clarification on how to address what Drakeo kindly pointed out above would be greatly appreciated!

unSpawn · 02-09-2008, 05:49 AM

Quote:

Originally Posted by tiber

If I go back and re-enable SELinux (enforcing), reload the LVM applet and the two volume groups magically show up.

Cool. So that's sort of evidence it's not due to recognition of your PERC-* but Something Else. The first thing IMHO to do would be to make the VG's visible the way you described and use the LVM tools to make certain all content is there and accessable since you didn't state you could.

Next it would be good to just get it over with: SELinux is here to stay and you should get used to it. Disabling it is only necessary in a minority of cases and the general reflex to disable it the instant it is perceived to cause problems isn't helping anyone. To cut things short: if you don't like it ponder moving to some distribution that isn't capable of offering extra layers of protection for free. So. Where from here? Most important thing is to collect information. A GNU/Linux system will have Syslogd log information to some files, check out /var/log for files like kernel, boot, messages and such. Check your last bootup log and see if there's anything there that requires attention. If unsure you could post excerpts here (preferably in BB code tags). If unsure what to post you could tarball them up, upload to some free hoster and post the D/L URI here. Another way to get information is to use commandline commands to perform system administration tasks (or start GUI apps from a terminal window if you can handle it any other way) because they could spit out relevant errors.

Only if you are sure your logs point to SELinux problems continue here. One problem you get when disabling SELinux is that your filesystems labelling will be out of sync. Issuing 'fixfiles restore' could help because it resets SELinux labelling to what's default. After that reboot stop Auditd, logrotate /var/log/audit/audit.log, reboot and see if things work. If it still doesn't then SELinux logs information into /var/log/audit/audit.log. These messages can be used to add rules to the SELinux policy which could help overcome problems that are local to your box. Try running these commands. If any errors occur post the contents of selinuxLocalPolicy.log.

Code:

selinuxLocalPolicy() { 
cd /etc/selinux/targeted/modules/active/ || { echo error stage0; exit 1; }
( cat /var/log/audit/audit.log; cat /var/log/messages ) | audit2allow -M local || { echo error stage1; exit 1; }
checkmodule -M -m -o local.mod local.te || { echo error stage2; exit 1; }
mv local.pp modules/local.pp || { echo error stage3; exit 1; }
semodule -i modules/local.pp || { echo error stage4; exit 1; } && \
 { echo "SELinux loading of local.pp finished OK."; cd; }; }
selinuxLocalPolicy 2>&1 | tee /tmp/selinuxLocalPolicy.log

tiber · 02-11-2008, 12:37 PM

Thanks for the reply unSpawn... I appreciate the time.

As I mentioned I'm not that well versed in Linux, but I agree from what I have seen that SELinux is likely here to stay for awhile. I would not have the slightest problem leaving it enabled, it's only the both Zimbra and Scalix want it disabled for proper installation and operation.
Scalix throws an error w/ SELinux enabled, and will not install until you clear it.

Zimbra will install, and while the system can be brought up it seems the MTA is not working.

I'll try to wade through the logfile info you mentioned, and see if I can make sense as to why the system doesn't like SELinux disabled or permissive. One would think that a fairly popular chipset (nforce4) would not cause these types of problems.

Thanks again,

Eddie