Bye bye OpenSolaris ?

So, I heard that OpenSolaris will be dead a.k.a discontinued.
Me my self, currently has no reason to use OpenSolaris. Why ? Simple question, what cant a Solaris box do compared to OpenSolaris ? currently Solaris also stil free, gcc stil runs on it, (sunfreeware.com ?). Need to install Oracle ? Oracle onlys supports Solaris distribution - not OpenSolaris. Need SUN's good old JDK ? use Solaris :) Not to mention that OpenSolaris distro's loves x86 too much instead Sparc ( well, almost all of my machines in my test lab were sparc :D ). Nexenta for example, the one that I -was- believe to be the ice breaker in Solaris world with their apt-get style, seems love x86 too much. Not too mention the stupidness of OpenSolaris AI installer for Sparc, which need an x86 OpenSolaris to be installed first, plus its additional macro thingy to be configured during wanboot. Why would I need that in Solaris ? Why cant I just plug the disc and do boot - install to my sparc machine ?

So now, there are only 2 options for my sparc machines, use Solaris, or Linux :D. Debian can run on Sparc platform. Just install sparc-utils, and we'll get the similar administration CLI just as we did on Sparc Solaris.

But too bad, Debian's apt package for solaris mostly compiled for 32 bit platform. If we need to boost the performace for our apps, we should re-compile it. No big deal, gcc is available :)

Solaris Jumpstart on Linux

Intro :

- Linux, Debian based (mine: Lenny), hostname : pegasus, act as the Solaris jumpstart server.
- Solaris, Solaris 10, hostname: solclient, will be the machine to be installed.
- IP addresses range 192.168.10.x (class C), make sure you use standard network class IP, this is where the standard broadcast address (192.168.10.255) comes into play during arp broadcast.
- solclient will broadcast arp request, replied by pegasus's rarpd, and then and IP address will be assigned. solclient will then download the minimum kernel using tftp. The kernel file name will be the ip address of the solclient. For example, if the assigned IP is 192.168.10.5, and then the kernel name will be C0A80A05. Another way is to use this command:

elinpra@pegasus > printf "%02X%02X%02X%02X\n" 192 168 10 5
C0A80A05

This file should be stored inside the root directory of tftp. Check the tftp setting on /etc/inetd.conf. Here is mine:

elinpra@pegasus > cat /etc/inetd.conf
tftp dgram udp4 wait nobody /usr/sbin/tcpd /usr/sbin/in.tftpd /srv/tftp/

Then, the file C0A80A05 should be put in /srv/tftp.

Steps :

1. Install bootparamd, rarpd, tftpd and nfs-kernel-server in Linux server (pegasus)
root@pegasus > apt-get install bootparamd rarpd tftpd nfs-kernel-server
2. Run iptables -F, make sure no firewall in Linux server
root@pegasus > iptables -F
3. Check the mac address of the solaris client.
Try boot from {ok}, see the mac address during booting (if the system is currently up, go to init 0 mode) :
ok boot net

Resetting ...
LOM event: +56d+21h39m43s host reset

þ
Sun Fire V120 (UltraSPARC-IIe 648MHz), No Keyboard
OpenBoot 4.0, 1024 MB memory installed, Serial #61911409.
Ethernet address 0:3:ba:b0:b1:71, Host ID: 83b0b171.

Back to lom> or sc> using 'break' command.

4. Set /etc/ethers and /etc/hosts in Linux server with the mac address and hostname of the solaris client.

elinpra@pegasus > echo "0:3:ba:b0:b1:71 solclient" >> /etc/ethers
elinpra@pegasus > echo "192.168.10.5 solclient" >> /etc/hosts

5. Start the rarpd if has not started yet.( I’m using debian based linux here )
root@pegasus > /etc/init.d/rarpd start

6. If you're installing using Solaris iso file, mount the iso file to a directory using the following command, otherwise if using Solaris CD/DVD, directly go to step 8:
root@pegasus > mkdir -p /mnt/sol10
root@pegasus > mount -o loop solaris_iso_file.iso /mnt/sol10
root@pegasus > cd /mnt/sol10

7. If you have the Solaris 10 CD, insert it on pegasus, do the following steps :
root@pegasus > cd /mnt/cdrom/ (to your solaris cdrom directory)

8. Find the suitable kernel for your machine type. Mine is SunFire v120, so it will be sun4u type. If your hardware is T5220, T2000 etc then you need sun4v kernel.
Locate the kernel on directory Solaris_10/Tools/Boot/platform/sun4u/inetboot for sun4u, or Solaris_10/Tools/Boot/platform/sun4v/inetboot for sun4v.
Copy it to /srv/tftp and make symbolic link to that file into C0A80A05.

root@pegasus > cp inetboot /srv/tftp
root@pegasus > ln -s /srv/tftp/inetboot /srv/tftp/C0A80A05

9. Start tftpd daemon in pegasus :
root@pegasus > /etc/init.d/openbsd-inetd restart

10. start nfs service in pegasus :
root@pegasus > /etc/init.d/nfs-kernel-server start

11. Prepare nfs mount so that solclient later on can download the entire installation package using nfs mount. Set the mount root into where the cdrom or iso files are mounted. Make sure that the nfs mount root from solclient is /sol10.

root@pegasus > exportfs \*:/mnt/cdrom -o fsid=0,ro,no_root_squash,crossmnt,no_subtree_check,sync <<- for CD/DVD or root@pegasus > exportfs \*:/mnt -o fsid=0,ro,no_root_squash,crossmnt,no_subtree_check,sync <<>Check if the nfs shares can be mounted correctly.

Playing around with low end VPS box

Last week I just bought a VPS account for playing around with. It feels so great having our own shell account with root login so that we can do anything inside. I bought it from here. First of all, I bought the minimalist one, with 64mb memory, with no burstable ram :D, using Centos. I simply cant do anything, their Centos installation itself took around 20-30mb ram. I also not well enough playing with Yum. This is not great, so I upgraded the account to 128mb ram. At the same time I noticed that besides Centos, the hosting provider also provide Debian Lenny as the OS....why dont you said that from the beginning ??? :D Now, I'm running Lenny with 128mb ram. The VM specs itself it's not that bad, considering it's powered with 2 cpu @3Ghz, 128mb ram with no swap ( this guy here shares a great script to create a fake swap, but I dont need it any longer ;) read the rest of the entry to know why... ), and with 10Gb disk, more than enough storage for minimalist like me :)

So, first thing first, chop all the unnecessary program, replace it with the program with smaller 'cost'. Just do ps -ef and there I found apache2 and Sendmail sitting around and eating my memory. Just do :

# apt-get remove --purge sendmail apache2
And dont forget to kill them, and remove them from startup script.
# kill -9 pid
# update-rc.d -f remove apache2
# update-rc.d -f remove sendmail
Ok, that left me around 5 mb used memory. Not bad :)
Next thing is to replace the default shell, since bash shell become greedy enough to eat my memory. I replaced it with pdksh since after googling around, this pdksh shell saves ~1mb for every shell session...good enough. So I install it :
# apt-get install pdksh
Dont remove the bash shell yet, since it has a lot of dependencies. Leave it there, but dont use it :) How do we do that ? Just edit your own default shell to become pdksh in /etc/passwd :
# vi /etc/passwd
Change the line using /bin/bash, into /bin/pdksh, and voillaa, next login session will be using pdksh.

Not enought with just # prompt ? It's time to decorate your shell with some fancy prompts, just edit the .profile located inside the user's home directory ( just do ls -la to see it, since it's a hidden file ). Add the following line :

HOSTNAME=`hostname`
PS1=`print "\027\033[1;32m$USER@$HOSTNAME $PWD > \027\033[0m"`

And then do logout and login again, now you have a green fancy prompt sitting around in your shell prompts.

Next thing we can chop with, is the rsyslogd daemon and the sshd daemon. I got a great reference here. We will replace the rsyslogd daemon with syslog-ng, and replace the open-ssh server with dropbear.
# apt-get install syslog-ng && dpkg --purge rsyslog
And then the dropbear.
# touch /etc/ssh/sshd_not_to_be_run
# apt-get install dropbear
And edit the /etc/default/dropbear to set NO_START=0. To make sure of what we have change, just reboot the server.
And after fresh reboot, here's what I have in my VPS :

root@ath.cx /root > ps -ef
UID PID PPID C STIME TTY TIME CMD
root 1 0 0 Apr05 ? 00:00:00 init [2]
root 340 1 0 Apr05 ? 00:00:00 /usr/sbin/syslog-ng -p /var/run/
root 348 1 0 Apr05 ? 00:00:00 /usr/sbin/dropbear -d /etc/dropb
root 367 1 0 Apr05 ? 00:00:00 /usr/sbin/xinetd -pidfile /var/r
root 382 1 0 Apr05 ? 00:00:00 /usr/sbin/cron
root 8864 348 0 02:24 ? 00:00:00 /usr/sbin/dropbear -d /etc/dropb
onty 8865 8864 0 02:24 pts/1 00:00:00 -pdksh
root 9192 8865 0 02:51 pts/1 00:00:00 -pdksh
root 9226 1 0 03:01 ? 00:00:00 nginx: master process /usr/sbin/
www-data 9227 9226 0 03:01 ? 00:00:00 nginx: worker process
root 9229 9192 0 03:01 pts/1 00:00:00 ps -ef

root@ath.cx /root > free -m
total used free shared buffers cached
Mem: 128 6 121 0 0 0
-/+ buffers/cache: 6 121
Swap: 0 0 0

Yep, I end up with just 6 mb ram for OS, dropbear, and nginx, and 121 ram free for my other apps.