Topic: Fantastic SME 7.3 virtual server installation running on Vmware/Centos 5.1/64 !

[arne]

The last 3 months or something like that I have been running and testing  SME 7.3 on Centos 5.1/64/Wmware server.

The host system is a minimum installation of Centos 5.1 and the free Vmware server is used for virtualization.

One of the really nice features of this virtualization alternative is the freedom og applying from 1 and up to 4 or even more network adapters without a problem, as these is supported directly from the underlying Centos 5.1 host system. I tested out the implementation of a Wireless access point into the Centos 5.1 host system. It worked, but I think it in some way "compromised the beauty of simplicity" of the Centos 5.1 host system, so I returned back to the option of using an external hardware wireless access point and only ordinary wired network adapters on the host system.

Then I am using a virtual Smoothwall firewall/gateway that can control the traffic for from and between from 1 to 4 network adapters using the easy graphical configuration tools of Smothwall Express. Behind the virtual gateway there is virtual SME 7.3 running as "server only". There is also a virtual Astlinux for the iptelephony server function, an extra virtual SME 7.3 server for testing purposes, a virtual Centos 5.1 server and a virtual Windows 2000 workstation for remote access and etc. (Reason for Win 2000 and not XP – unused old licenses and less server load.)

Even though there is some other stuff on it I still think on it and consider it to be "my SME server" as the functions that I am using on daily basis is based at the "virtual SME 7.3 production server". (Not real production, but rather a lab/test setup in a private home.)

The idea for the installation has been to keep the basic host system completely "clean and unmodified". Then also to keep the SME 7.3 production server "clean and unmodified".  All testing of contribs and alternate stuff is done on the virtual test server.

Virtual servers that is not in use or not needed at the moment is closed down, and then activated when needed.

When it comes to backup options the ordinary SME bacup tools and methods can be used for backup, or the whole and complete virtual server image can be stored on some local storage media or it can be transfered via network. I think that a few days ago, for some testing, I transfered a Vmware image of the size of 6-7 GB between the Centos host system and a Windows XP workstation using SSH and WinSCP. It wasn't a problem at all.

The experience of testing out this concept, "the virtual sme 7.3 on Centos 5.1" can be summed up in two words: "stability and flexibility".

There is not a technical issue on anything, any more, and there is also almost no limitations at all, how things can be set up, and how things can work.

Approx 3 months of testing have until now given 0.0 bugs. It was just set up and then it was running.

It has happened a few times that the virtual SME 7.3  test server have been messed up, but then it has been just to replace it with the test server backup image. The ability to test out almost anything without affecting the stability of the production server or production gateway, might be the greatest advantage of this alternative.

Hopefully there is nothing controversial about the virtualization alternative. The SME server works very well together with this technology both as a virtual client and as a virtual host system.

[cactus]

Approx 3 months of testing have until now given 0.0 bugs. It was just set up and then it was running.

If so, please tell me how you solved the time sync of your SME Server OS on VMWare server as I still have not found the proper method...

[arne]

At least I have not noticed any errors.

I think I remember it was something with time sycronization .. but what ..??

I just now checked the time of the underlaying host operating system and the time of the virtual sme server. They are exact syncrone.

Do you have some more description of the problem .. (to fresh up the memory) .. ?

[cactus]

At least I have not noticed any errors.

I think I remember it was something with time sycronization .. but what ..??

I just now checked the time of the underlaying host operating system and the time of the virtual sme server. They are exact syncrone.

Do you have some more description of the problem .. (to fresh up the memory) .. ?

See page 11 (Linux section) of this link: http://www.vmware.com/pdf/vmware_timekeeping.pdf

[m]

cactus,

I have experienced time drift issues a long time ago, where I use the following cron job as a workaround

Code: [Select]

*/1 * * * * root (/usr/sbin/ntpdate pool.ntp.org || (/sbin/e-smith/service ntpd stop; /usr/bin/svc-stop ntpd)) > /dev/null 2>&1

On all my current virtual SME 7 installations the problem has disappeared (I guess with SME7/Kernel 2.6) and the time is correct w/o the cron job.

[arne]

A shorter text than the one I first posted here:

Syncronization of the host system and the guest system apear to be dependent on the installation of the software package vmware-tools on the virtual server. The reason that is was syncronized here was because I had this ad-on on the virtual sme server.  http://kbase.redhat.com/faq/FAQ_43_9259.shtm

While testing I observed that other virtual servers that can not run vmware-tools, Astlinux and the virtual Smoothwall gateway, had a time setting that drifted away. I also observed that the host system did not have any syncronization with an external time server.

The ntpd time syncronization mechanism was installed for the host system: "yum install ntp"

Service were then set up to start up automatically using "ntsysv". Service was also started and restartet using "service ntpd start/stop/restart".

After that I made an initially manual time configuration on host system and all virtual Linux server using the date command "date 04051430" to ease a later automatic syncronization. For virtual Smoothwall gateway and the Astlinux the time parameter was only set this way as they does not support "vmware-tools". (Explanation of the use of the date command 04=April, 05=date 5, 14=hours 14, 30=minutes 30)

After restarting the ntpd on the host system, syncronization of the host system with external time servers vere obtained.

After then restarting ntpd on the virtual Linux servers with Vmware-tools installed and, they syncronised as well.

Now the host system and the virtual installations should be syncronised, but it is reason to believe that the virtual Smoothwall gateway and the virtual Astlinux iptelephony server will drift away after some time, so they will have to be manually adjusted, later on, using the date command. Reason for this is that they don't have the Vmware-tools package installed.

I used this link when working on the time syncronization issue: http://www.linuxhomenetworking.com/wiki/index.php/Quick_HOWTO_:_Ch24_:_The_NTP_Server

I believe that everything should work ok with proper time syncronization, for all of it, just now.

[arne]

For an automated timesyncronization of the virtual Smoothwall gateway it apeared to be possible to use a variant of mweinber mentioned above.

The Smoothwall has a ntpd installation, that should not be running. Then this command is implemented as a (1 hour ?) cron job:

"ntpdate -u 10.0.0.99"

where 10.0.0.99 is just an excample of an ip belonging to the host system.

Looks like this method will give a time syncronization that is precise enough for the most practical purposes for virtual systems that has ntpd but not vmware-tools installed.

The firewall guest will then be syncronizing with the host and the host will have to be syncronized with an external time server.

An interesing aspect of this problem is that the host is actually doing its external time syncronization via the nat function of its virtual gateway guest (in this case Smoothwall). Because of this mechanism, it is not possible to run the host system with zero ip adresses. From a security point of view, zero ip adresses might be intersting, alternative, but then there also will be zero external time server syncronization for the host system.. (??!)

man ntpdate : http://linux.die.net/man/1/ntpdate

cactus -> Thanks a lot for mentioning the time syncronization problem. I have to admit I had not noticed that it initially was sat up as poor as it was. Now it should be OK.

[arne]

New reviced text:

After a few days, of testing, I found out that the time syncronization worked fine for the host system and for all guest systems exept for the Astlinux installation. I were not able to correct this error and desided to test out alternatives.

Tree different alternatives vere desidet to tested out: 1. To implement a Asterisk rpm to the virtual SME 7.3 server. 2. To make a virtual installation of the new complete AsteriskNow server from Digium. 3. To set up a minimum virtual Centos 5.1 and install Asterisk via rpm's or from source code.

The alternative that apeared not to work at all were the new server from Digium. It were not willing to be installed at all as a virtual server. Don't know why.  http://www.asterisknow.org/ (When googleing I were only able to find prebuilt 64 bit vmware images and no 32 bits images. Might be some problems related to hardware.)

One other alternative that did not work were to install Asterisk on a virtual Centos 5.1 from rpm's. I were not able to find rpm's that worked.

Then the one of two alternatives that seems to be working without a problem:

Downloaded and installed the Selintra Asterisk rpm's for sme server and installed this to the virtual SME 7.3. No problems I beliveve, but not the newest updated software. http://distro.ibiblio.org/pub/linux/distributions/smeserver//contribs/selintra/RPMS/

As the virtual SME server is running already with a proper time syncronization, the Asterisk server running on this one should do it as well.

Then it was also to do the alternative of a Asterisk server installed from source code on a "minimum" Centos 5.1:


1. Made the basic minimum virtual installation over internet, using the net installation CD.

2. Downloaded VMwareTools-1.0.4-56528.i386.rpm (Don't know where I found it at the Vmware webpage)

2. Installed rpm: rpm -ivh VMwareTools*

3. Added DAG/Rpmforge repository:
rpm -Uhv http://apt.sw.be/redhat/el5/en/i386/rpmforge/RPMS/rpmforge-release-0.3.6-1.el5.rf.i386.rpm
(Not actually neccessary, this time, but could be nice to have for later use.)

4. Yum update

5. Yum install gcc

6. Yum install kernel-devel*

7. Yum install ntp

8. Runned: vmware-config-tools.pl 

Basic server were up and running with vmware tools, time syncronization etc.

I first tried to install Asterisk from rpm's but it did not work.

I then found this howto on the net for an Asterisk installation from sourcecide:

http://www.asterisktopics.com/?p=14

It apeared that the software links were a bit old so I used new links:

http://ftp.digium.com/pub/libpri/releases/
http://ftp.digium.com/pub/zaptel/releases/
http://ftp.digium.com/pub/asterisk/releases/

Then just followed the howto.

There were not any bigger issues to make the source code installation and the application came up and is running. (Using 20 times more space than the Astlinux for the same job.)

Looks like it is now running two different alternatives, that both has the proper time syncronization and that both can do the job, asterisk running at the virtual SME server, installed from rpm's and Asterisk running at the virtual Centos 5.1 installation, installed from source code.

[Stefano]

After testing it out a few days, I found out that the time syncronization worked fine for the host system and for all guest systems exept for the Astlinux installation. This is just a very minimalistic Asterisk installation, and I'm not able to find the right way to make time syncronization for this one.

Hi..

you only say that "it did not work"..

it's a bit OT but can you explain the problem?

a quick search on google with "astlinux ntpd update" gave me many results

ciao
Stefano

[arne]

Thanks for the attention. Astlinux is my real favorite Asterisk server variant so, I would prefere it to be the alternative no 1.

The problem is that there is, as far as I know only two ways to make an virtual Linux server running with proper time syncronisation:

1. To install Vmware tools. As this also require a compilator, I think this is out of the question for Astlinux that is really a minimalistic Linux distro.

2. To run the ntpdate command from a cron job, and I think that the Astlinux neither has a cron job (or has it ?) or a ntpdate command (It only said something like "unknown command" when tested from the ommand line.)

The situation is just that the time function is just drifting away, while the two ways of doing the syncronization (I know about) will not work (as far as I know.) (I am also running a Asterisk server on ordinary hardware and this does not have a problem with time syncronization at all.)

If the time syncronization could be done, while running as a virtual server, I would prefere to use Astlinux again.

 

[Stefano]

2. To run the ntpdate command from a cron job, and I think that the Astlinux neither has a cron job (or has it ?) or a ntpdate command (It only said something like unknown command when tested from the ommand line.)
 

according to this http://mirror.astlinux.org/userguide026.pdf

AstLinux 0.2 has the following features (in no order):
• Linux kernel 2.6.11.7
• dnsmasq
• watchdog (busybox)
• ssmtp
• cron (busybox)
• bash
• busybox (and everything with it...)
• TFTP server
• vsftpd FTP server
• mini_httpd (with SSL)
• PHP in CGI mode (patched for above)
• OpenSSH
• OpenSSL
• NTP (ntpdate, ntpq, ntpd)
• syslog (busybox with remote log support)

this is for astlinux 0.2

HTH

Ciao

Stefano

[arne]

Thanks. Maybe I did not check it well enough. I will try to give it a new try. A small problem with the Astlinux is that it in som way is built up a bit "unstandard", so things are often not found where you expect to find them.

From testing with ntpd on Astlinux:

pbx ~ # service ntpd stop
Stopping ntpd...
Terminated
pbx ~ # service ntpd start
Starting ntpd...
pbx ~ # ntpdate -u 10.0.0.99
-sh: ntpdate: command not found
pbx ~ #

My version of Astlinux: 0.4.8


At the moment it looks like that the Asterisk rpm installation at the sme smeserver and the sourcecode installation on the dedicated virtual Centos 5.1 both works with proper time syncronization. If anyone know how to make the Astlinux to run the proper way as a virtual server with time syncronization and everything working, it would be great.

AND FOR SECURITY REASONS IT MIGHT BE A GOOD IDEA TO REMOVE COMPILATORS ETC WHEN FINISHED.