Opus Interactive | If These Servers Could Talk

When I was asked to complete my VMware Sales Professional (VSP) certification through their partner central I was not sure what to expect. But by the time I was finished, I was perplexed by VMware’s capabilities, as well as the advantages that they can offer to the companies that use it and to the ones that should be.

 

I came to realize through my  training that VMware is the recognized leader in virtualization technology solutions and they have certainly earned the title. After some time in front of prospects, I began to understand that I actually had the upper hand when sitting down with them. Because I completed my VSP I truly understand what VMware is able to offer to its  partners and customers.   Not only is the software itself incredible but the VMware Partner Network provides companies with the tools they need to gain a long-term competitive advantage with virtualization.

 

Knowing and understanding the products that your company uses is essential to the success of any business, and VMware couldn’t make it easier for its partners to do so. If VMware and virtualization is something you’re unfamiliar with, it’s time to get up to speed. You’ll be pleased you did.

When you reach a point where you no longer have enough free space on your server and simply adding a new drive and symlinking to it will not really suffice, it’s time to grow the filesystem. When I first skimmed over websites explaining how to do this, it seemed overly complicated. Alll you really need to do is a couple of calculations and then use those numbers as the new parameters.

Modifying the filesystem can be dangerous, so make sure you have a backup of any data you do not want to lose. Luckily, I was working on a virtual server, so I just had to clone it, which takes a little time, but less painful than rebuilding the server and restoring from backup.

The first step is to see the partition table as it currently is:

 root@web01:~ > fdisk -s
 /dev/da0: 3394 cyl 255 hd 63 sec
 Part        Start        Size Type Flags
    1:          63    54524547 0xa5 0×80

 root@web01:~ > fdisk
 ******* Working on device /dev/da0 *******

 parameters extracted from in-core disklabel are:
 cylinders=3394 heads=255 sectors/track=63 (16065 blks/cyl)

 Figures below won’t work with BIOS for partitions not in cyl 1
 parameters to be used for BIOS calculations are:
 cylinders=3394 heads=255 sectors/track=63 (16065 blks/cyl)

 Media sector size is 512
 Warning: BIOS sector numbering starts with sector 1
 Information from DOS bootblock is:
 The data for partition 1 is:
 sysid 165 (0xa5),(FreeBSD/NetBSD/386BSD)
     start 63, size 54524547 (26623 Meg), flag 80 (active)
  beg: cyl 0/ head 1/ sector 1;
  end: cyl 1023/ head 254/ sector 63
 The data for partition 2 is:
 <UNUSED>
 The data for partition 3 is:
 <UNUSED>
 The data for partition 4 is:
 <UNUSED>

The current partitions are:

 root@web01:~ > df -h
 Filesystem     Size    Used   Avail Capacity Mounted on
 /dev/da0s1a    248M    166M     62M    73%    /
 devfs          1.0K    1.0K      0B   100%    /dev
 /dev/da0s1d    989M    8.3M    902M     1%    /tmp
 /dev/da0s1f     22G     18G    2.3G    88%    /usr
 /dev/da0s1e    1.9G    302M    1.5G    17%    /var
 procfs         4.0K    4.0K      0B   100%    /proc

 root@web01:~ > disklabel /dev/da0s1
 # /dev/da0s1:
 8 partitions:
 #        size   offset    fstype   [fsize bsize bps/cpg]
   a:   524288        0    4.2BSD     2048 16384 32776
   b:  1048576   524288      swap
   c: 54524547        0    unused        0     0         # “raw” part, don’t edit
   d:  2097152  1572864    4.2BSD     2048 16384 28552
   e:  4194304  3670016    4.2BSD     2048 16384 28552
   f: 46660227  7864320    4.2BSD     2048 16384 28552

Here we see that /usr is 46660227 sectors, or ~22G. The size can be checked with the formula:
 (46660227/2)/1024 = 22783M
 
I am adding another 20G of space, reverse the above formula to figure out the size:
 42*1024*1024*2 = 88080384 sectors
 
We can use the final number of sectors to now determine the final size of the entire filesystem:
 (88080384-46660227)+54524547 = 95944704 sectors

We also need to figure out how many cylinders will be used, (16065 sectors/cylinder):
 95944704/16065 = 5972
 
The summary of changes to be made (and adjustments I had to make to account for inaccuracies in the formulas translating to the actual disk structure):

 cylinders: 3394 -> 5972 (6004 this change was made automatically by the system)
 c: 54524547 -> 95944704 (95940117 this was adjusted by fdisk)
 d: 46660227 -> 88080384 (88075797 this was figured by subtracting the same number of sectors that fdisk dropped to properly align the cylinder boundary)
 
I had to make those adjustments to avoid getting error messages after running fdisk.

**At this point you will need to log into single user mode and then mount /tmp

 # mount /tmp
 # fdisk -u
 ******* Working on device /dev/da0 *******
 parameters extracted from in-core disklabel are:
 cylinders=6004 heads=255 sectors/track=63 (16065 blks/cyl)

 Figures below won’t work with BIOS for partitions not in cyl 1
 parameters to be used for BIOS calculations are:
 cylinders=6004 heads=255 sectors/track=63 (16065 blks/cyl)

 Do you want to change our idea of what BIOS thinks? [n]
 Media sector size is 512
 Warning: BIOS sector numbering starts with sector 1
 Information from DOS bootblock is:
 The data for partition 1 is:
 sysid 165 (0xa5),(FreeBSD/NetBSD/386BSD)
     start 63, size 54524547 (26623 Meg), flag 80 (active)
  beg: cyl 0/ head 1/ sector 1;
  end: cyl 1023/ head 254/ sector 63
 Do you want to change it? [n] y
 Supply a decimal value for “sysid (165=FreeBSD)” [165]
 Supply a decimal value for “start” [63]
 Supply a decimal value for “size” [54524547] 184025088
 fdisk: WARNING: partition does not end on a cylinder boundary
 fdisk: WARNING: this may confuse the BIOS or some operating systems
 Correct this automatically? [n] y
 fdisk: WARNING: adjusting size of partition to 184024512
 Explicitly specify beg/end address ? [n] y
 Supply a decimal value for “beginning cylinder” [0]
 Supply a decimal value for “beginning head” [1]
 Supply a decimal value for “beginning sector” [1]
 Supply a decimal value for “ending cylinder” [1023] 5972
 Supply a decimal value for “ending head” [254]
 Supply a decimal value for “ending sector” [63]
 sysid 165 (0xa5),(FreeBSD/NetBSD/386BSD)
     start 63, size 95940117 (46845 Meg), flag 80 (active)
  beg: cyl 0/ head 1/ sector 1;
  end: cyl 852/ head 254/ sector 63
 Are we happy with this entry? [n] y
 The data for partition 2 is:
 <UNUSED>
 Do you want to change it? [n]
 The data for partition 3 is:
 <UNUSED>
 Do you want to change it? [n]
 The data for partition 4 is:
 <UNUSED>
 Do you want to change it? [n]
 Partition 1 is marked active
 Do you want to change the active partition? [n]

 We haven’t changed the partition table yet.  This is your last chance.
 parameters extracted from in-core disklabel are:
 cylinders=3394 heads=255 sectors/track=63 (16065 blks/cyl)

 Figures below won’t work with BIOS for partitions not in cyl 1
 parameters to be used for BIOS calculations are:
 cylinders=3394 heads=255 sectors/track=63 (16065 blks/cyl)

 Information from DOS bootblock is:
 1: sysid 165 (0xa5),(FreeBSD/NetBSD/386BSD)
     start 63, size 95940117 (46845 Meg), flag 80 (active)
  beg: cyl 0/ head 1/ sector 1;
  end: cyl 852/ head 254/ sector 63
 2: <UNUSED>
 3: <UNUSED>
 4: <UNUSED>
 Should we write new partition table? [n] y
Then I rebooted and choose single user mode again:

 # fdisk -s
 /dev/da0: 6004 cyl 255 hd 63 sec
 Part        Start        Size Type Flags
    1:          63     95940117 0×06 0×00
 #mount /tmp
 #mount /usr
 #disklabel -e -r /dev/da0s1

 8 partitions:
 #        size   offset    fstype   [fsize bsize bps/cpg]
   a:   524288        0    4.2BSD     2048 16384 32776
   b:  1048576   524288      swap                       
   c: 54524547        0    unused        0     0         # “raw” part, don’t edit
   d:  2097152  1572864   4.2BSD     2048 16384 28552 
   e:  4194304  3670016    4.2BSD     2048 16384 28552
   f: 46660227  7864320    4.2BSD     2048 16384 28552
 #

Here we change:

 c: 5452547 -> 95940117
 f: 46660227 -> 88075797

Next we unmount /usr and will use growfs to grow it.

 #umount /usr
 #growfs -s 88075797 /dev/da0s1f

Once growfs is finished, mount /usr again and check your partitions with df -h:

 # mount /usr
 # df -h
 Filesystem    Size   Used  Avail Capacity  Mounted on
 /dev/da0s1a   248M   166M    62M    73%    /
 devfs       1.0K   1.0K     0B   100%    /dev
 /dev/da0s1d   989M   8.3M   902M     1%    /tmp
 /dev/da0s1e   1.9G   302M   1.5G    17%    /var
 /dev/da0s1f    41G    18G    20G    47%    /usr
 #

Then we can run fsck to make sure the drive is ok. Once you are done, reboot in regular mode and you are done.

The world of Hosting has been a crazy roller coaster ride of technology shifts over the past 10 years through the dot-com bust and the current economic landscape.  We saw that before the dot-com bust, and even again about a year to a year and a half ago before the current economic downturn, that overly complex and sometimes unused technologies are chosen by small startup companies that are not needed.  The expected load and traffic for these customers is often over-estimated, leading to hundreds of thousands and sometimes millions of wasted investor dollars being spent when not necessarily needed.  Large Java applications that are custom written (re-inventing the wheel in most cases) with large Interwoven CMS systems, massive Application Layers using JBOSS and custom classes for payment vendor gateway integrations were being developed.  The resulting hosting solution was often underutilized and segmented due to the application design and not optimized for ultimate scalability.
 
As the hosting technologies are changing and evolving into what is now a mainstream viable platform (cough-cough “cloud”) we hosters can provide a solution that allows the flexibility and scalability to treat the hosting platform as a true utility service – which will help with the current economic downturn (I’ll discuss this more in my next blog posting).  Whether it is grid or cloud computing, the offerings are numerous and the verbiage on how it gets described is usually close but stated differently in the market across the providers.  The hosting industry is heading towards the utility hosting service – much like electricity, natural gas and water.  You can use what you need, when you need it and only pay for your usage – a true utility model.  We saw this trend heading this direction before Green IT was over marketed and before grid computing and “cloud” computing was being pushed as it is today and created our grid offering vGrid and our scalable on-demand VMware hosting platform vClustr.  We started offering this back in early 2007 and have continued to see very strong growth in these offerings even today. 

Up next are details on why we continue to see and expect strong growth in spite of today’s economic forecast.

We have recently built a large Virtual Cluster (vClustr) here at Opus Interactive (I am still trying to get management to name it the “HerbaCluster”) and I thought I would write a couple of blog postings about how it is built and why we chose the vendors we did.

 

The first thing we had to do was decided on the Virtualization platform. We started off by testing the front runners in Virtualization – XenSource, Virtual Iron, Microsoft Virtual Server, Microsoft Hyper-V and VMware. There were a number of things that we were looking for when doing our testing. The first requirement was that the software had to run on HP’s bl460c Blade Servers. The c-class blade servers are the servers that we have standardized on here at Opus Interactive, so it was important that we could use this hardware with whatever platform we chose. After testing the different Virtualization platforms came decision time. Needless to say it was not an easy choice to make, but in the end we went with VMware.

 

The cost of VMware was a big deterrent but you could not beat the features that VMware had to offer.  VMware is the “king” in the virtualization space right now and we wanted to take advantages of all the features that VMware offered. We purchased numerous VMware’s ESX 3.5 Server licenses along with a VMware VirtualCenter license. This gave us the ability to take advantage of such features as HA (High Availability), DRS (Dynamically Allocate System Resources), VMotion, Update Manager and Consolidated Backup.

 

HA (High Availability) and VMotion are my favorite features right now (things like Consolidated Backup you don’t fall in love with until you get woken up in the middle of the night because a Virtual Machine is complete hosed). The ability to be able to move Virtual Machines between physical servers with nothing more than one or two dropped pings is a dream come true for a System Engineer working in a datacenter. All the sudden, tasks that would require long periods of down time for a customer is no longer a concern because those customers can be migrated to another server while things like firmware are upgraded, memory is installed or additional processors are installed. Some late night maintenance windows can now be completed during normal work hours (I mean business hours, we all know that System Engineers don’t know what normal work hours are) and with HA (High Availability) you don’t have to worry that a hardware problem is going to take down a large number of customers – if an ESX Server goes down, HA will migrate all of the Virtual Machines to different ESX Servers minimizing down time even more!

 

DRS (Dynamically Allocate System Resources) is something that will become more important as more Virtual Machines are added to this environment (currently the cluster is so large that there is no way that any Virtuals could have problems contending for resources). If at some time in the future the virtual Machines begin to need more resources then the physical server can provide, DRS will move the Virtual Machine to a physical server that has more resources available. This feature is great for customer who experience unusual traffic (perhaps things featured on Slashdot or a holiday website). One of the great things about DRS is that it is not just some feature that you turn on. DRS lets you setup criteria that has to be met before it will migrate the Virtual Machine to a new host. This leaves me more control over what is going on in the cluster and keeps Virtual Machines form bouncing around from host to host.

 

Update Manager and Consolidate Backup are features that are a requirement for any IT system. Patching and backups are part of everyday tasks that have to be completed and it is nice to have these tools come packaged with everything else.

by Jeremy Sherwood

Opus Interactive has teamed with the Portland Lumberjax in its quest to return to the Championship game.

The Portland Lumberjax, the city’s professional lacrosse team, is entering its fourth year in the league and is hoping to return to the Championship game, where they were edged out by the Buffalo Bandits, 14-13. The team made an improbable run to the finals last year.

With lacrosse being one of the fastest growing sports in the United States, we saw this chance to work with the  West Division Champions as an opportunity we couldn’t pass up. As a long time lacrosse player, coach and season ticket holder, I couldn’t be happier to help support the sport and the team. I had a great oppurtunity over the last couple of days to meet with Cristi and Jessica (Jax Dancers) and 

Brodie Merrill, Pat Jones and Tyler Codron to tour the Opus Interactive Data Center.  To see more photos from our tour click here. These three players and dancers are a prime example of why we are proud to sponsor the Lumberjax.  All of them give of their time and talents to the city of Portland. They spend time at schools, hospitals, and charitable organizations, giving back to the city in which they play.  I don’t know of very many professional athletes that not only work at their sport and passion, but also work other jobs and still find more than enough time to give back.

In turn, we saw a great opportunity to bring our technology expertise to the Lumberjax hosting and email environments. We saw this as a great chance for us to give back as well. “Anyone who has been to our website knows it had gotten a bit, we’ll say sloppy, over the past few years.  So, when the folks from Opus Interactive approached me about upgrading our systems I was very intrigued.  I quickly went from intrigued to impressed when we had our first meeting and toured their facility.  Opus opened my eyes to internet and website possibilities, security and backup, that I had never dreamed of,” stated Brian Silcott, Executive Vice President at Portland Lumberjax.

One of the best ways many of us can support the Portland Lumberjax is by buying tickets and watching the games.  You can buy ticket here, and see some of the live action for yourself. We hope to see you there.