Note: The VM will need to be shutdown or suspended (actually, running the following command will suspend the VM anyway):

For Xen Server 5.5 update 1 and later:

coalesce-leaf –u <uuid of VM>

For Xen Server 5.6 and later:

xe host-call-plugin host-uuid=<host-UUID> plugin=coalesce-leaf fn=leaf-coalesce args:vm_uuid=<VM-UUID>

A small script (kudos to http://sysadminnotebook.blogspot.com/2011/07/reclaim-disk-space-from-deleted.html) to automate the process if you need:

dd if=/dev/zero bs=100M | gzip | gzip -d | gzip | gzip -d | gzip | gzip -d > /dev/null &

Because of the non paged memory pool limit of 256MB on the Windows 32 bit systems, I wasn’t able to use Virtuozzo for Windows on dedicated systems with more than 8GB. This translated into a large volume of work related to build and deploy the dedicated systems. In addition, around 2GB of memory was never being use which is a waste.

The solution:

Being the fact that Virtuozzo is not a hypervisor, working mostly like a patch to the operating system, I’ve got the idea of running Virtuozzo as a virtual machine under one of the popular hypervisors. First I try Xen opensource for which Virtuozzo installer crashed with an error related to some virtual device, then I wanted to try Citrix XenServer, but I dropped the idea because it’s still based on Xen opensource and I was in a hurry.

The winner was Vmware with its ESXi 4. This one performed excelent so now I can squeeze four Virtuozzo for Windows 32 bit  in a 24GB RAM system – easier to manage and better use of resources (disk and memory).

So, here is a local link for the qmhandle tool.

Shutdown domU, increase the disk space for domU using XenCenter, insert a rescue CD (Gentoo would do the job), attach it to domU and restart domU in Recovery Mode.

After booting:

- using fdisk, delete the LVM partition from the disk

- recreate the partition. The new partition will use the additional disk space

- change the partition type to Linux LVM, save and quit

- resize the physical volume (pvresize)

- reactivate the volume group (vgchange -a y vol_group_name)

- resize the logical volume (lvresize)

- resize the filesystem (resize2fs)

That’s all, we’re back in business.

Of course, if you can’t afford the downtime, then you can simply add a second virtual disk for this domU and use the LVM tools increase the disk space.

Plesk autoinstaller -

CentOS 5 32 bit: http://download1.parallels.com/Plesk/PPP9/CentOS5/parallels_installer_v3.6.0_build100407.15_os_CentOS_5_i386

CentOS 5 64 bit: http://download1.parallels.com/Plesk/PPP9/CentOS5/parallels_installer_v3.6.0_build100407.15_os_CentOS_5_x86_64

Virtuozzo 4.6 for Windows:  http://www.parallels.com/download/pvc46/win/

Virtuozzo 4.5 for Windows: http://www.parallels.com/download/pvc45/win/

Virtuozzo 4.o link has disappeared from Parallels site, but the installer still work and you can get it from here (add proper exe file extension manually).

Google-ing a bit on this matter I came across this post in Cacti forum – http://forums.cacti.net/about25600.html – pointing to an open source tool called snmptools. The tool can be downloaded from http://erwan.l.free.fr/snmptools/ and it works perfect on my Virtuozzo servers. Now I can read the Non Paged Memory Pool value using Zabbix and get a notification when the system is about to reach the limit. The next step will be to read the value in Cacti and send the notification using Thresholds plugin for Cacti.

On the second server (the destination, meaning the one that the LV will be copied to):

root@server2 ~# nc -l IP_address port_number | dd of=/dev/VolGroup/LV_name

On the first server (the source, meaning the one that LV will be copied from):

root@server2 ~# lvcreate -L size -n LV_name VolGroup

root@server2 ~# dd if=/dev/VolGroup/LV_name |nc IP_address port_number

If the source server is in production, you may want to start the copy process with nice (or renice later) in  the order to avoid a high I/O stress on the system.

There is no progress bar, but you can confirm that the copy operation work fine by checking the system activity with “top” or by quickly check the network traffic with “tcpdump”.  At the end, if no errors are returned, you’ll have the same LV on both systems. One more thing. In the above example the source LV was not being in use. If your system is using the LV, better creates a snapshot of it and use the snapshot for the copy operation.

It may sound like a trivial task – “What could be so complicated to chose a dedicated server? Just go online, search for some good feedbacks of a few hosting companies, then chose whichever fits your needs.” The reality is a bit different and is related to the competition in this market.

The common word is “price”. In today’s global economy situation (and crisis I may add), big companies as well as small companies are affected by their costs related to the hardware used for the dedicated servers they provide to their customers. In order to be competitive, they all have to keep their costs down.  Any hosting company site you’ll visit advertise the same thing and this is “server”.

Wikipedia provides the following definition of the  server computer: “A server computer, sometimes called an enterprise server, is a computer system that provides essential services across a network, to private users inside a large organization or to public users in the internet. [...] Enterprise servers are known to be very fault tolerant, for even a short-term failure can cost more than purchasing and installing the system.”

What the above definition means? That the hosting company are always using an enterprise level server, fault tolerant and so on? That may be the general understanding but I assure you that’s not the case.  Actually, in the real life, most of the time, for cheap dedicated servers plans, all you get is actually a desktop motherboard packed in the best case in a 1U box so that it will fit nicely in a rack and the company will get a better space utilization in their datacenter. For those who don’t know the difference between a desktop and a server class hardware, here are some differences.

The server class hardware is designed to run for years without shutting down the computer. A desktop class hardware is designed for home/office use and the average daily usage would be 16 hours. This is reflected in the parts quality used for motherboards, hard drives, memories, power supplies. Server class hardware is being roughly testes and the tests are more comprehensive then for the desktop class hardware. Server class components have a mean time between failures (MTBF) parameter which basically means how much time can you expect to pass between 2 consecutive failures and it represents a guarantee of the quality of the component. Desktop class hardware doesn’t come with such parameter.

Let’s talk about some practical examples. From the server class motherboards manufacturers we can mention Dell, HP, Intel, Supermicro. I named just a few and I’m not saying that are not other very good manufacturers of this type of hardware. For example, an Intel SR1530AH is coming with Intel S3000AH motherboard (chipset Intel 3000) and it has a high quality power supply delivering 650W. The case looks very good and it comes with a easy to install pair of rails. Assuming that the box is equipped with an Intel Celeron D440, 1GB of RAM and 80GB SATA drive, the price at an Intel integrator may vary between $600-$800US.  The same configuration except for the motherboard, the case and the power supply may cost $250-$300US (for example a CybertronPC 1U box with Asus P4N-MX motherboard and a 250W power supply with a low cost low quality CPU fan and a few other fans spread in the box to compensate the lack of the air flow because of the components placed on the motherboard.

For some hosting companies, it makes a lot of sense to limit the initial costs so guess what… they will chose a low cost type of hardware. What will be the differences? The chances that something will brake are higher for a desktop class hardware than for a server class hardware.

The memory technology are actually pretty good these days. You’ll not see often failures here. It’s not the same for the motherboard or for the hard drives. Motherboards include electrolytic capacitors. There are different qualities and, obviously, low cost hardware will always use low cost (read low quality) capacitors. They are mostly used around CPU area and they have a critical influence over the stability of the system. A CPU for which the power supply is not properly filtered or stable will determine random errors which may translate in file systems corruption, sudden death of the system (it just reboots or shutdown without any other warnings). People which held an used, older motherboard may have noticed that some capacitors are slightly “ballooned” or ever worse, cracked in the upper are. This is a clear indication of cheap stuff which failed in time because of the electrical and temperature stress.

Another example is the hard drives. There are cheap desktop hard drives on the market that sells for less than $60US for a 80GB SATA drive. There are chances that such drive will fail. It may also work just fine for a few years. You may see the same type of hard drive,  the same manufacturer, but different lots. One lot will fail, one not. The problem is that you can’t determine before using if it will fail or not and which one.

The next on the list is the power supply. Modern power supplies are switching-mode power supplies. As most of the electronic devices they use capacitors (see above) as well as switching transistors ((bipolar or MOS-FET) and magnetic cores. A low end power supply will deliver poor performances affecting the system stability and performances.

Now, if you know all these but the downtime or data loss is not an issue for you, then you don’t have to worry. But if you expect a high uptime then you can gather some information from the hosting company sales people before deciding if it’s a good system or not. If you’re not a technical person, you may ask a friend to help here.

So, here are a few questions that you can ask the sales people:

- who is the manufacturer of the server advertised in the hosting company plan and which model it is? Once you have this information, you can simply check on the manufacturer website and see what kind of hardware you’re about to pay for. There is one more catch here. You can get a real server but one that is not manufactured anymore. A typical warranty these days is 3 years for a server. It may go beyond this limit, but you’ll pay for the warranty extension. There is a reason for a manufacturer to give a default limited time warranty. It means that it believes that the server will be ok for that period of time and not more. If the server is a “no name” one but you are still interested in this hosting company, ask about the manufacturer of the motherboard or at least about the chipset from it. If you can’t get a proper answer it means that either they are using crap or their technical knowledge is really limited. Personally I find more dangerous a hosting company with non-technical people (yes, they are, you’ll be amazed) then a low quality hardware. This is because a highly trained technical person will always be able to help you in any situation, including hardware failures, while a non-technical person will only succeed to increase your level of frustration, eventually after spending more money.

- who is the manufacturer of the hard drives and which models are they? There are desktop class hard drives and enterprise class hard drives. Ask about MTBF of the drives. Again, if you can’t get proper answers, my advice would be to stay away.

- power supply – if the server is a real server class then the power supply will definitely be a good one. If not, they will probably not be able to give you anything more than the power of the power supply, measured in Watts. Something under350W should raise a brow especially if you have more than one hard drive and the CPU is not a low power one (the last generation of CPUs are using less power than the previous generation, but even in the last generation, there are CPUs using more power than others).

- server case – the real servers have good cases – for example, for Intel SR1530 a production engineer from Intel told me they spend 4 million US dollars for the design.  A cheap server box will be affected by the vibrations from the rack – don’t forget that your server runs in a rack with another 41 servers (if we’re talking about the 42U racks) so vibrations are inherent. Vibrations transmitted to cheap fans and hard drives increase the chances of a component failure.

No matter which hosting company you’ll chose, be sure that you have backups of your data. Don’t rely on the hosting company backups if you don’t pay for it. Normally a backup feature “looks good” in advertising. If they advertise a backup plan including in your dedicated server plan, ask for details like the backup solution used, the backup policy, the backup size limits, then, after a first month, ask for some data to be restored. Do this from the beginning and you’ll avoid a lot of trouble later.

Well, that would be all for now. Please feel free to add a comment or to contact us if you have any questions.

So we have a domain, is being hosted on a server to a hosting  provider and we want to  send some emails.  It may sound a  simple task, but these days, with  so many  protections against  spam, we  can easily  found  our messages being delivered directly in other people’s junk folders. However, there are few  things that  can drastically  improve the  chances for  the message to hit the right spot – the recipient inbox folder. Here they are.

First, be sure that the IP address of your server has a proper reverse DNS record. Most of the providers will allow you to add your own rDNS record. If not, you’ll have to ask their technical support to do it for you.

For example, if your domain is domain.com and your IP address is 123.234.123.234, be sure that the IP address will resolve to some hostname from your domain. It could be just the domain name. Most of the time you’ll see something like mail.domain.com.

[root@server ~]# host -t mx domain.com

domain.com mail is handled by 10 mail.domain.com.

[root@server ~]# host mail.domain.com

mail.domain.com has address 123.234.123.234

[root@server ~]# host 123.234.123.234

234.123.234.123.in-addr.arpa domain name pointer mail.domain.com.

Second, give a  name from  your domain to  your email  server. Some  email servers default installations use the generic name “localhost.localdomain” and this is a very  bad idea when it  comes about sending emails.  Sending messages with this identity will determine your messages to be rejected by most of the receiving email servers. Please note that not all the time the system name will be  the email server name  (for example Qmail with  Plesk control panel). You can simply check if your server is properly configured from this point of view by following the few steps described here.

Third, add a proper SPF record to your domain name zone file. A good  page about this can be found here. The information from this page will  help you to define your SPF record in no time. Also use DomainKeys and DKIM (DomainKeys Identified Mail) in order to identify your email server as the legitimate sender of the emails for your domain.