au boulot

Installing openSSH on Solaris 8

2008-09-11T10:42:00.002+02:00

Nice page describing how to install/setup openSSH on Solaris 8 at :
http://sunfreeware.com/openssh8.html

Installing a linux zone on a Solaris host

2008-09-05T18:03:00.004+02:00

Previous posts describe the creation of Solaris zones, this post describes the creation of a linux zone (RedHat EL 3.x or CentOS 3.x) running on a Solaris 10 host (x86 hardware).

First create a ZFS filesystem for your zone as described in the previous posts on this blog.
Here we will use the zfs file system zones_pool/zone1_root.

Then create the zone :

zonecfg -z linux-zone1
linux-zone1: No such zone configured
Use 'create' to begin configuring a new zone.
zonecfg:linux-zone1> create -t SUNWlx
zonecfg:linux-zone1> set zonepath=/zones_pool/zone1_root
zonecfg:linux-zone1> set autoboot=true
zonecfg:linux-zone1> set bootargs=""
zonecfg:linux-zone1> add net
zonecfg:linux-zone1:net> set address=10.17.7.156/24
zonecfg:linux-zone1:net> set physical=bge0
zonecfg:linux-zone1:net> end
zonecfg:linux-zone1> verify
zonecfg:linux-zone1> commit
zonecfg:linux-zone1> exit

To install linux on the zone, you need one linux distribution (today RedHat 3.x or CentOS 3.x),

either as a tar archive,
or a set of cd-rom/dvd,
or a set of isos images.

You can for example download a CentOS tar archive from :
http://opensolaris.org/os/community/brandz/downloads/

Here, we will install RedHat from isos images, the zoneadm installation command is given in parameter a directory where to find these iso files :

zoneadm -z linux-zone1 install -d /net/machine/export/LINUX/as3-u8/i386_iso

Installing distribution 'Red Hat Enterprise Linux 3 Update 8'...
Installing cluster 'desktop'
Installing miniroot for zone 'linux-zone1'.
Attempting to locate 30 packages...

Installing 1 miniroot package...

Attempting to locate 29 packages...

Installing 29 miniroot packages...

Setting up the initial lx brand environment.
System configuration modifications complete.
Duplicating miniroot; this may take a few minutes...
Booting zone miniroot...
Miniroot zone setup complete.

Performing full install for zone 'linux-zone1'.
Attempting to locate 663 packages...

Installing zone 'linux-zone1' from ISO
"/net/machine/export/LINUX/as3-u8/i386_iso/RHEL3-U8-re20060712.1-i386-disc2-ftp.iso".

Checking for deferred packages...
Attempting to locate 3 packages...

Installing 470 packages; this may take several minutes...

Attempting to locate 193 packages...

Installing zone 'linux-zone1' from ISO
"/net/machine/export/LINUX/as3-u8/i386_iso/RHEL3-U8-re20060712.1-i386-disc3-ftp.iso".

Checking for deferred packages...
Attempting to locate 3 packages...

Installing 106 packages; this may take several minutes...

Attempting to locate 87 packages...

Installing zone 'linux-zone1' from ISO
"/net/machine/export/LINUX/as3-u8/i386_iso/RHEL3-U8-re20060712.1-i386-disc4-ftp.iso".

Checking for deferred packages...
Attempting to locate 3 packages...

Installing 69 packages; this may take several minutes...

Attempting to locate 18 packages...

Installing zone 'linux-zone1' from ISO
"/net/machine/export/LINUX/as3-u8/i386_iso/RHEL3-U8-re20060712.1-i386-AS-disc1-ftp.iso".

Checking for deferred packages...
Attempting to locate 3 packages...

Installing 18 packages; this may take a few minutes...

Installing 3 deferred packages; this may take a few minutes...

Completing installation; this may take a few minutes.
Setting up the initial lx brand environment.
System configuration modifications complete.
Installation of Red Hat Enterprise Linux 3 Update 8 to zone
'linux-zone1' completed Fri Sep 5 17:33:06 MEST 2008.

Installation of zone 'linux-zone1' completed successfully.

Details saved to log file:
"/zones_pool/zone1_root/root/var/log/linux-zone1.install.20498.log"

By default, networking is disabled in a linux zone.
You can enable it by editing from the Solaris global zone, the file /zones_pool/zone1_root/root/etc/sysconfig/network to have the following content:

NETWORKING="yes"
HOSTNAME=myhost
NISDOMAIN=mydomain

You can now boot the zone, using the command:
zoneadm -z linux-zone1 boot

Installing a Solaris 8 zone on CMT hardware

2008-07-25T15:38:00.003+02:00

Solaris 8 is not supported on new systems like the Sun Fire T2000 using the Chip Multithreading Technology (CMT), but it is possible to install a Solaris 8 zone on this system and run a Solaris 8 application.

For this example, we will use a Sun Fire T2000 running Solaris 10 5/08.
(for solaris 10 8/07, the instructions are the same, but you need first to download from sunsolve and install patch 127111-01 or the latest patch that obsoletes 127111)

First download Solaris 8 Containers 1.0 from https://cds.sun.com/is-bin/INTERSHOP.enfinity/WFS/CDS-CDS_SMI-Site/en_US/-/USD/ViewProductDetail-Start?ProductRef=solaris8-cont-1.0-G-F@CDS-CDS_SMI
(packages and the sample Solaris 8 flash archive)

Install two of the packages contained in this download :

pkgadd -d . SUNWs8brandr
pkgadd -d . SUNWs8brandu

Create a ZFS filesystem for your zone as described in the previous post on this blog.
Here we will use the zfs file system zones_pool/zone1_root.

Then create the Solaris 8 zone and specify the machine hardware name being sun4u (while our physical system using UltraSPARC T1 is a sun4v hardware) :

zonecfg -z s8-zone
s8-zone: No such zone configured
Use 'create' to begin configuring a new zone.
zonecfg:s8-zone> create -t SUNWsolaris8
zonecfg:s8-zone> set zonepath=/zones_pool/zone1_root
zonecfg:s8-zone> set autoboot=true
zonecfg:s8-zone> add net
zonecfg:s8-zone:net> set address=10.17.7.249/24
zonecfg:s8-zone:net> set physical=e1000g0
zonecfg:s8-zone:net> end
zonecfg:s8-zone> add attr
zonecfg:s8-zone:attr> set name=machine
zonecfg:s8-zone:attr> set type=string
zonecfg:s8-zone:attr> set value=sun4u
zonecfg:s8-zone:attr> end
zonecfg:s8-zone> exit

Install the zone from the flash archive sample provided by Sun in the Solaris 8 Containers 1.0 download :

zoneadm -z s8-zone install -u -a /download/solaris8-image.flar
Log File: /var/tmp/s8-zone.install.23636.log
Source: /download/solaris8-image.flar
Installing: This may take several minutes...
Postprocessing: This may take several minutes...

Result: Installation completed successfully.
Log File: /zones_pool/zone1_root/root/var/log/s8-zone.install.23636.log

You can now boot the zone, login on this zone, and check the output of uname command :

uname -a
SunOS s8zone 5.8 Generic_Virtual sun4u sparc SUNW,Sun-Fire-T200

Creating a zone with a root directory on a ZFS file system

2008-07-17T16:59:00.004+02:00

Basic try of commands to create a zone on a Solaris 10 8/07 host.

First create a zfs file system where we will later store our zone root.
Our host has just one disk, so for this test we will create a zfs pool from a 8Go partition of this disk that already contains a UFS filesystem , but it can be overridden using the zpool create -f option:

zpool create -f zones_pool c0t0d0s3

Check it was created:

zpool list
NAME SIZE USED AVAIL CAP HEALTH ALTROOT
zones_pool 7.94G 88K 7.94G 0% ONLINE -

Create a zfs filesystem:

zfs create zones_pool/zone1_root

Check it was created:

zfs list
NAME USED AVAIL REFER MOUNTPOINT
zones_pool 115K 7.81G 25.5K /zones_pool
zones_pool/zone1_root 24.5K 7.81G 24.5K /zones_pool/zone1_root

Set a quota on this filesystem (not mandatory, better for isolation, and it can be changed easily with zfs):

zfs set quota=800m zones_pool/zone1_root

Change the permissions as the zone root must not have read/write/execute permissions for group and other:

chmod go-rwx /zones_pool/zone1_root

Our zone root filesystem is ready, we can now create a zone:

zonecfg -z zone1
zone1: No such zone configured
Use 'create' to begin configuring a new zone.
zonecfg:zone1> create
zonecfg:zone1> set zonepath=/zones_pool/zone1_root
zonecfg:zone1> add net
zonecfg:zone1:net> set physical=bge0
zonecfg:zone1:net> set address=10.19.1.22/24
zonecfg:zone1:net> end
zonecfg:zone1> exit

Install this new zone:

zoneadm -z zone1 install
Preparing to install zone .
Creating list of files to copy from the global zone.
Copying <547> files to the zone.
Initializing zone product registry.
Determining zone package initialization order.
Preparing to initialize <445> packages on the zone.
Initialized <445> packages on zone.
Zone is initialized.
The file contains a log of the zone installation.

Check the size of our zone's root:

zfs list
NAME USED AVAIL REFER MOUNTPOINT
zones_pool 11.8M 7.80G 25.5K /zones_pool
zones_pool/zone1_root 11.7M 788M 11.7M /zones_pool/zone1_root

This doesn't take much place as the zone we created is a sparse root zone: only a subset of the packages installed in the global zone are installed in our new zone, other packages are accessed as read-only loopbak mounted filesystems.

Another option is to create a whole root zone where all the required packages are installed into the zone's root. This is done by removing default inherit-pkg-dir resources.

We can see the default inherit-pkg-dir resources of our zone by running the command:

zonecfg -z zone1 info
zonename: zone1
zonepath: /zones_pool/zone1_root
brand: native
autoboot: false
bootargs:
pool:
limitpriv:
scheduling-class:
ip-type: shared
inherit-pkg-dir:
dir: /lib
inherit-pkg-dir:
dir: /platform
inherit-pkg-dir:
dir: /sbin
inherit-pkg-dir:
dir: /usr
net:
address: 10.19.1.22
physical: bge0

Now we have to configure the zone.
This can be done either in interactive mode or by creating a configuration file.

For the interactive mode, login on the console:

zlogin -C zone1

And from another terminal, boot the zone:

zoneadm -z zone1 boot

A series of questions about language, locale, etc... will appear on the console.

Or we can configure the zone in non-interactive mode by creating a sysidcfg file:

create a file /zones_pool/zone1_root/root/etc/sysidcfg with the following content:

system_locale=C
terminal=dtterm
network_interface=primary {
hostname=hostzone1
}
timeserver=localhost
security_policy=NONE
name_service=NONE
timezone=MET
system_locale=C
root_password=""
nfs4_domain=dynamic

Then boot the zone:

zoneadm -z zone1 boot

Login on the console zone:

zlogin -C zone1

Check the host name:

hostname
hostzone1

Check the network configuration:

ifconfig -a
lo0:1: flags=2001000849 mtu 8232 index 1
inet 127.0.0.1 netmask ff000000
bge0:1: flags=1000843 mtu 1500 index 2
inet 10.19.1.22 netmask ffffff00 broadcast 10.19.1.255

We just see the network configuration of zone1 and not the network configuration of the global zone

Type ~. to leave zone1 console

Disable ipv6 on RedHat 4

2008-04-17T14:25:00.002+02:00

Edit the file /etc/modprobe.conf and add the following line :
alias net-pf-10 off

Then reboot the node.

TIPC on Solaris

2008-03-28T16:05:00.001+02:00

TIPC (Transparent Inter-Process Communication) is now available on your preferred OS !
With Renaud Métrich, we ported TIPC on Solaris.
Packages for Solaris 10 and later releases for both SPARC and x86 can be downloaded from sourceforge or from OpenSolaris.

Creating a patch on linux

2008-03-28T15:56:00.001+02:00

diff -Naur -x exclude_file path_1 path_2 > patch_name

rcp/rsh fails on error rcmd: socket: Cannot assign requested address

2008-03-28T15:49:00.005+02:00

On one of our Solaris 10 systems, it happened from time to time that a rsh or rcp command was failing during the NHAS installation, with the following error:

rcmd: socket: Cannot assign requested address

This is a solaris 10 problem that has been addressed by the following bug :

6533773 tcp checksum 0xFFFF (-0) used instead of 0x0000 (+0) since Solaris 10

A Patch is available on sunsolve to fix this issue : patch 127111

Running OpenClovis on top of NHAS

2007-07-09T13:55:00.000+02:00

The Netra HA Suite (NHAS) provides High Availability foundation services like:

a Reliable IP Transport, CGTP (Carrier Grade Transport Protocol). It ensures that a single network failure does not result in data loss and provides zero time fail-over IP.
an implementation of the Service Availability Forum (SA Forum ) Cluster Membership Service (CLM) API triggering fast fail-over, isolation of faulty node and performing cluster membership.
a Reliable NFS Service, providing a reliable file system accessible through NFS, enabling fail-over of NFS data and services to backup in case of failure.
a Reliable Boot and DHCP Service for diskless nodes, enabling node reboot regardless of potential hardware or software failures in the cluster.
etc,...

OpenClovis is a open source carrier-grade middleware providing among many features :

an implementation of the SA Forum Availability Management Framework (AMF) allowing to configure and manage redundancy models for HA components (2N, N+1, ...).
a checkpointing service,
an implementation of the SA Forum Event Service,
a Group Membership Service (GMS) detecting when new nodes join or existing nodes leave the system, and providing a cluster view and cluster membership notification service to components subscribing for the service.
etc, ...

This Group Membership Service (GMS) provided by OpenClovis is using some piece of code from the third party OpenAIS middleware :

TOTEM, a reliable multicast protocol,
the CLM, cluster Membership API, modified for GMS needs.

Replacing this OpenAIS piece of code with calls to the NHAS APIs would provide OpenClovis features like checkpointing, or the Application Management Framework of 2N/N+1 redundancy availabilty models, on a cluster running NHAS foundation services.

The details of code modifications to get a prototype of OpenClovis running on top of NHAS, or on top of any third party software providing an implementation of the SA Forum Cluster Membership API, are described in the document: Using OpenClovis on top of NHAS.

Parallelization in perl

2007-04-13T18:44:00.000+02:00

Some piece of perl code we are using to execute operations on a set of cluster nodes, is doing the job sequentially node by node.
To speed up things, I had a look on how to parallelize this piece of code in perl.
Perl provides threads and fork.
But threads support is a compilation option, so it may not be provided on all platforms.
To know if the current perl version is compiled to support threads, you can do :

$ perl -V | grep thread
usethreads=undef use5005threads=undef useithreads=undef usemultiplicity=undef

Here, the perl version bundled with Solaris 10 doesn't provide threads feature.
So I tried to use fork, and it's working fine, here is a sample of code.
There was function func taking one parameter that was called n times with a different parameter.
The function below parallelize each call to this function func :


sub execute_in_parallel {

 my ($func, @param_list) = @_;

 my $pid;
 my @children;
 my $final_status;
 
 foreach my $param (@param_list) {

  if (!defined($pid = fork)) {

   print "ERROR: parent could not fork\n";
   exit(1);

  } elsif ($pid) {

   # We are in the parent
   # Store the child PID
   push(@children, $pid);

  } else {

   # We are in a child
   # do the job
   my $res = &$func($param);
   exit($res);

  }
 }

 # Wait for the end of children
 $final_status = 0;
 
 foreach my $childId (@children) {

  waitpid($childId, 0);
  my $status = $?;
  if ( $status != 0 ) {
   $final_status = $status;
   print "child running $func exited with status $status\n";
  }

 }
 
 if ( $final_status != 0 ) {
  print "One child exited with status $final_status";
 }

}

Installing a Sun host from its system console

2007-04-05T10:07:00.000+02:00

From a Sun Netra T2000 system console, a power-on will boot the host.
if instead of a boot, we want to perform an OS installation from the network, the default auto-boot parameter has to be changed from the system console as follows :

sc> poweroff -yf
sc> bootmode bootscript="setenv auto-boot? false"
sc> poweron -c

We will be connected to the console. We will then have to wait for the prompt ok and type :

ok setenv auto-boot? true
ok boot net - install

First steps in ruby

2007-04-04T14:18:00.000+02:00

I had a quick look at ruby for some internal tools, the language looks really nice with a lot of extensions.

As I'm using a cluster made of several nodes in my job, I wrote a first little program to launch in parallel some Netra HA Suite admin commands on each node to get the cluster membership status, trigger a switch-over, etc... or run any system command.

I removed the Netra HA Suite specific stuff here and just let the code running any system command in parallel.
This code is using the NET::SSH ruby extension.
And it assumes that ssh is running on cluster nodes and configured to allow root login without password.
First, launch the ruby intepreter (irb) with the lib cluster.rb provided at the end of this post:

$ /usr/local/bin/irb -r cluster.rb --simple-prompt --noinspect
>>

You are on the interpreter, now you can define your cluster and specify the hostnames of nodes part of your cluster :

>> c = cluster.new
>> c.add 'v20z-1'
>> c.add 'v20z-2'
>> c.add 'v20z-3'

Run a system command on all nodes, date for example to see if nodes are synchronized :

>> c.run 'date'
v20z-1 : Fri Mar 23 18:53:03 GMT 2007
v20z-2 : Fri Mar 23 18:53:01 GMT 2007
v20z-3 : Fri Mar 23 18:53:05 GMT 2007

No they aren't, synchronize them :

>> c.run 'date 03231852'
v20z-3 : Fri Mar 23 18:52:00 GMT 2007
v20z-1 : Fri Mar 23 18:52:00 GMT 2007
v20z-2 : Fri Mar 23 18:52:00 GMT 2007

Check again to see :

>> c.run 'date'
v20z-1 :Fri Mar 23 18:52:15 GMT 2007
v20z-2 :Fri Mar 23 18:52:15 GMT 2007
v20z-3 :Fri Mar 23 18:52:15 GMT 2007

Here is the ruby file cluster.rb :


require 'rubygems'
require 'net/ssh'

# Global variables

$user='root'
$passwd='none'

######################################################################
#
# Runs a command given in parameter
# Parameters : a ssh session and the command to run
#
def run_cmd( session, cmd )
   
    # open a shell

    shell = session.shell.sync
       
    # Get the hostname
   
    out = shell.hostname
    myhost = out.stdout.chomp

    # Run the command
   
    out = shell.send_command cmd
    status = out.status
    result = out.stdout.chomp
    if (status != 0)
        puts "ERROR on #{myhost} : #{result}"
    else
        puts "#{myhost} : #{result}"
    end
   
end

######################################################################
#
# cluster class to instantiate by the user
#
# Provides the methods :
#
# . add        : to add a node in the list of cluster nodes
# . run        : to run a system command in parallel on all nodes
#
class cluster

    attr_accessor :nodes
   
    def initialize
        @nodes = Array.new
    end
   
    # Add a node to the list in the cluster
   
    def add(node)
        @nodes.push(node)
        self
    end       

    #
    # Run in parallel on all node a command given in argument
    #
   
    def run(cmd)
   
        if @nodes.nil?
            puts "No node in the cluster yet..."
        else

            # For each node, start a new thread
            # Each thread will :
            # - Start a new SSH session for the node
            # - run the command
           
            threads = []
            @nodes.each do |node|
           
                threads << Thread.new(node) { |myNode|
               
                    Net::SSH.start(
                    myNode,
                    $user,
                    $passwd ) do |session|
                   
                    res = run_cmd(session, cmd)
                end
               
                }
               
            end
           
            # Wait for the threads to finish
           
            threads.each { |aThread|  aThread.join }
        end
    end

end

opensolaris jumpstart installation

2007-04-04T11:18:00.000+02:00

During a re-installation of a host with the latest Solaris 10 update 4 build or opensolaris, if the jumpstart installation is unexpectedly interactive you can:

check if the following new parameter is defined in the sysidcfg jumpstart file:
nfs4_domain=dynamic

check if the time service running on the install server, and if not, enable it doing :
$ svcadm enable svc:/network/time:dgram
$ svcadm enable svc:/network/time:stream