Study Notes

I’ve not received my RHCE certification for a week. This issue makes me so anxious since this certification is very very important to me. As Red Hat official reply, they will track the examiner to submit me archieve. Yes, I feel confident to pass the difficult exam, but long waiting is hard time.

Install Windows after Linux installation.

Fixed an old problem this week. When I try to install Windows in Linux box which has been install Red Hat Enterprise Linux 4.4, select boot device from CD-ROM, the Windows installation program only shows a few text and system seems to be hang. First, I thinks it is cause by CD corrupt and change another Windows CD, it has the same symptom. I googled with this issue and found nothing userful. Last time I had to add second harddisk and install new grub, which took a very long time. When thinking of this, perhaps it’s causes by MBR locked. Hmm, things come to easy. “dd if=/dev/zero of=/dev/hda bs=1024 count=512” to clear all MBR information on the harddisk. Of course original Linux is not able to start again for missing MBR info, but now, Windows operating system would be installed successfully. According to the fixing process, MBR stores operating system boot info and partition table in the first 512 byte of harddisk. It is effective to clear MBR in order to get pure environment.

Connect Linux boxes with Serial port

During tuning kernel, we need to check information when system boots. For example, there’re two Linux machines,
192.168.123.143
192.168.123.67
We want to redirect all 143 information to 67. This could be done by following steps:
1. Login 143, modify /boot/grub/grub.conf
title 2.6.19
root (hd0,0)
kernel /vmlinuz-2.6.19 ro root=LABEL=/1 console=ttyS0,19200
initrd /initrd-2.6.19.img
2. Also in 143, modify /etc/inittab, add the marked line.
# Run gettys in standard runlevels
S0:2345:respawn:/sbin/agetty -L 19200 ttyS0 vt100
3. Restart 143, press DEL to enter CMOS setting,
Console Redirect —- 19200, VT100, COM1, Enabled.
4. Login 67, run minicom program,
# minicom
CTRL+A – Z – O – minicom setting – Serial Port – 19200 8IN 1, ttyS0
Save as default setting and restart minicom.
# minicom -C /root/kernel.log
At this point, you would see all the information of 143 and save the output to /root/kernel.log.

his is the first article of 2007. Through one month preparing, I take the RHCE exam today in International Culture Build. I got 90 scores in the morning part, and feel confident to pass the second Installation/Configuration exam for finishing all the jobs within 2 hours(3 Hours restricted). During the passed 40 days, less-sleeping, pratice-practice-practoce, read a lot of manuals and guides, I think I am able to win this battle for my hard working.

Yesterday, completed reading the RHCE(7.3) documents, and started up RHCE(3.0) compaign.
Plan to 3 days. The next one is RedHat Manual.

What time is the best to take RHCE exam? Today, I’m really annoyed this problem, since I decided to get RHCE cert in Octorber. In recent two months, I insist on reading Study Guide every night, and do not finish the course now. Sometimes I feel anxious and a litter disapointed.

I call some exam companies this afternoon, who told me an exam arranged in 29th Jan, 2007 and the fee is 4300 excluding training cost. I think it’s too expensive to have such a training only for rapid passing exam(Training+Exam=8100 in Shenzhen, and 6000 in Guangzhou). If join such “rapid-training camp” and even pass RHCE very easily. is it really benifit for me? Do I only need such an paper to declair how clever I am, or just depend on it to get a better job? Do I really master the study points of Linux world?

Absolutely not. It wastes my money and violates my rule to become a Linux professioner. I perfer to take more time to perpare, read more carefully, do more experiences, and recite the key knowledges by writing them on papers. Till all is ready and I feel confident, I directiva take exam without joining training. It will save me a lot of money and make my heart more stronger.

Go ahead, with more patience.

It tooks about two days on resolving how to update kernel 2.6.19.1 in VMware5.5, because the VMware default configured disk is SCSI which need some special module setting. During this course, I learn more about kernel concepts.

1. Target
Update the lastest kernel 2.6.19.1 in VMware5.5 in order to research NFSv4.0

2. Test Environment
VMware version: 5.5.1 build -19175 for linux
Virtual Operating System: CenOS 3.5 (Kernel 2.4.21-32EL)

3. Steps
Install Virtual Operaing System CenOS3.5 which kernel is 2.4.21. Download the kernel 2.6.19.1 with related patch from www.kernel.org.
# cd /usr/src
# tar xvf /download/linux-2.6.19.1.tar
# ln linux-2.6.19.1 linux-2.6
# cd linux-2.6

Apply the patch:
# bunzip2 -dv /download/patch-2.6.19.bz2 | patch -p1

Begin to compile:
# cd /usr/src/linux-2.6
# make mrproper
# make menuconfig (make oldconfig would be ignored here)
At this point, the following items must be selected as “build in”, more explanation would be wrote in the later section.
SCSI device support <*>
SCSI device -> SCSI low-lever device -> Buslogic SCSI support <*>
Device Drivers -> Fusion MPT sevice support <*>
Device Drivers -> Fusion MPT (base+scsihost) drivers <*>
Device Drivers -> Fusion MPT (ioctl) drivers <*>
Device Drivers -> USB support <*>
Device Drivers -> Network support -> Ethernet(10/100) support <*>
File System -> EXT2, EXT3 support <*>
Commend, the vmware default SCSI controller is Buslogic.

If not add “scsi, usb, mpt” built in directly, or just select modules load [M], when issue “mkinitrd /boot/initrd-2.6.19.1.img 2.6.19.1”, there’s the following error message:
no modules mptbase found for 2.6.19.1 kernel, aborting
If not add “ext2, ext3” build in directly, or just select modules load [M], when reboot after compiling, the system will be halt:
Kernel Panic: no init found

Continue the compiling process:
# make bzImage
# make modules
# make modules-install
# cp /usr/src/2.6.19.1/arch/i386/boot/bzImage /boot/vmlinuz-2.6.19.1
# mkinitrd /boot/initrd-2.6.19.1.img 2.6.19.1
# cp /usr/src/2.6.19.1/System.map /boot/System.map-2.6.19.1
# cp /usr/src/2.6.19.1/.config /boot/config-2.6.19.1

Modify “/boot/grub/grub.conf”:
# This is Phillip’s /boot/grub/grub.conf
default=0
timeout=10
splashimage=(hd0,0)/grub/splash.xpm.gz
title CentOS-3 (2.4.21-32.EL)
root (hd0,0)
kernel /vmlinuz-2.4.21-32.EL ro root=LABEL=/
initrd /initrd-2.4.21-32.EL.img
title New-Kernel (2.6.19.1)
root (hd0,0)
kernel /vmlinuz-2.6.19.1 ro root=LABEL=/
initrd /initrd-2.6.19.1.img
Save and reboot, if all succeed, issue “uname -r” will shows: 2.6.19.1

4. Summary
If set hard disk as IDE, it would be very easy to update. But I think practise on this SCSI system is very benifit for future works. At first begining, when issue “mkinitrd /boot/initrd-2.6.19.1.img 2.6.19.1” and error “no modules mptbase found for kernel”, I just did a stupid thing, add the “–omit-scsi-modules”(ignore scsi), but this system is based on SCSI drivers! Though “mkinitrd –omit-scis-modules /boot/initrd-2.6.19.1.img 2.6.19.1” passed, the system was kernel panic when reboot:

VFS: Cannot open root device “sda1” or unknown-block(0,0)
Please append a correct “root=” boot option
Kernel panic: not syncing: VFS: unable to mount root fs on unknown block(0,0)

By reading some documents, I tried to comment all lines contains “mpt” in /etc/modules.conf, the same kernel panic happenes; Download the latest “lvm2, device-mapped, mkinitrd” rpm packages, however, these packages can not be installed. The only way fixed this problem is building mpt,scsi,ext2,ext3,usb directly when run “make menuconfig”.

menuconfig options:
[*] build-in
[M] modules
[ ] excluded

5. Fix issues in new 2.6 kernel
The network device eth0 seems to be disappeared in new kernel, when execute “#lsmod”, there’s the following error:
QM_MODULES: Function not implemented.

I fix this issue by installing new mode-init-tools(need download),
# cd mode-init-tools
# ./configure –prefix
# make moveold
# make
# make install
Generate “/etc/modprobe.conf” :
# ./generate_modprobe.conf /etc/modprobe.conf
# depmod 2.6.19
OK, reboot the OS, and issue “#lsmod” again, the eth0 module “pcnet32” is listed.(vmware default eth0 module is called pcnet32). Fixed.

In some cases, I need to log in as remote X client, for example, run ConsoleOne in localhost.

Method 1:
System1# ssh -X system2
System2# run any X command you’d like have and it’ll show up on System1
I verifies this way to open remote ConsoleOne GUI(192.168.123.33, nds server) in local(192.168.123.32).
Localhost 192.168.123.32# ssh -X 192.168.123.33
NDS sev 192.168.123.33# ./usr/ConsoleOne/bin/ConsoleOne

Note, that is that does not work it’s because you have X11 forwarding turned off in your ssh config file(perhaps in /etc/ssh/sshd_config).

Method 2:
System1# xhost -system2
System1# telnet system2
System2# export DISPLAY=system1:0
System2# run X command you’d like have and it’ll show up on System1

–display Server:0.0
Send output to the Xserver running on Server.

Xterm is an X-client application that creates a terminal windown on your X-display

Mobo: the abbreviation of MotherBoard.

In my virtual CentOS, try to apply its OpenLDAP installed. First, check “/etc/openldap/slapd.conf”, it used “ldbm” instead of Berkeley DB “bdb”, set the base DN, password and thus configuration is completed. I add a lot of entries for validation, and later delete them. Note, here the destination in “ldapdelete” command must like the following:

# ldapdelete -x -v -D “cn=root,dc=plasmon,dc=sit” -w hello123 uid=inter,ou=People,dc=plasmon,dc=sit

Before the term “uid=inter” there’s no “dn:”, if such title exists, the ldapdelete would get failed.

I want to browse Internet pages via lynx, but now lynx do not support gb2312 directly. So, I did a very interesting test, resolve this issue fairly well.

1. Test Environment
CentOS 3.5, no install X-Window
cce rpm package

2. Recompily kernel to increase the running speed(because it has PIII and only 128MB memory).

# cd /usr/src/linux-2.4.21-32.EL
# vi Makefile
Here, in the line 4, change the EXTRAVERSION to “EXTRAVERSION=-32.ELhx”
# make mrproper
# make oldconfig
# make menuconfig
remove the unuseful components
# make dep
# make bzImage
# make modules
# cp /usr/src/linux-2.4.21-32.EL/arch/i386/boot/bzImage /boot/vmlinuz-linux-2.4.21-32.ELhx
# make modules_install
# mkinitrd /boot/initrd-2.4.21-32.ELhx.img 2.4.21-32.ELhx
# cp /usr/src/linux-2.4.21-32.EL/System.map /boot/System.map-2.4.21-32.EL
# cp /usr/src/linux-2.4.21-32.EL/.config /boot/.config-2.4.21-32.ELhx

Final step, add corresponding lines in /boot/grub/grub.conf. Then reboot.

3. The new kernel starts up very quickly, That’s what I want. Download cce and install this rpm file.
# rpm -ivh cce-xxx.rpm
Then execute “cce” in the command line, and type “lynx www.tom.com”, a famous Chinese site.
Yeah!I got it, all Chinese characters display very beautifully. cce also supports Chinese input, but what types can not be recognized by human for mess code set.

PS: copy cce package from windows to my Linux, and issue “rpm -ivh”, it said the md5sum failed. Download cce in linux, and installed successfully. It seems there’s different md5 checking code in Windows and Linux. I can not confirm.

It sounds to be very annoyed to import posixAccount, shadowAccount attributes and objectclasses in last article. Now I find a easy way. Located the “rfc2307-usergroup.sch” file in “/opt/novell/eDirectory/lib/nds-schema”, issue:

# ndssch -h 192.168.123.33 -t Plasmon cn=PlasmonAdmin.o=Plasmon rfc2307-usergroup.sch

The job would be done well. and then just use ConsoleOne to import “samba-nds” as rename it to ldif file.

To get the OpenLDAP schema entries:
# ldapsearch -x -D “cn=root,dc=plasmon,dc=sit” -w hello123 -b cn=subschema -s base + -H ldap://192.168.123.33:389
In OpenLDAP slapd.conf, it was set as “cn=subschema”, while in Iplanet, it is “cn=schema”.
To get the Iplanet schema entries:
# ldapsearch -x -b cn=schema -s base+ -H ldap://192.168.123.33:389

Assume that do not know the LDAP server type, issue the following command to find it is “cn=schema” or “cn=subschema”:
# ldapsearch -x -H ldap://192.168.123.33:389 -s base -b “” “objectclass=*” subschemaSubentry

Iplanet history: Iplanet 4.0 — Iplanet 5.0 — Sun One Directory Service — Sun Java System Directory Service 5.2