Wright PC Consulting, LLC

Update: More Down time for Surpass/HostDime.

Early in the morning on Friday, May 23, 2008 power was lost at the HostDime data center that houses Wright PC Consulting, LLC’s leased server. While, that is not that big of a deal, there are both UPS and a Generator for such occasions, what followed is. For some it meant 16 hours or more of downtime.

Here is what happened and what is being done to prevent such future failures.

To our clients and business partners,

There are no words to describe how deeply we apologize about the downtime which occurred on Friday, May 23, 2008. The incident has created immense discontentment to our organization mentally and emotionally because of the love and dedication our team has to our entire community. Moreover, because we realize the level of damage this incident has potentially caused you. We know there is neither money nor words which will replace the losses that may have been experienced by each one of you. Our organization is forever in debt to you all for the frustration and grief endured. It is never easy in disasters, but many of you showed your support as we worked non-stop to get things back to normal. We want to thank all of you for your patience, understanding, and support during such a difficult time. In any case, a formal incident report of our investigation is what we wish to rightfully deliver to you. Below is the detailed summary of events as they occurred. Please note some of you may have not experience any outage during this, not all clients were effected but we wanted to keep everyone updated.

What happened:

At approximately 8 A.M. EST our data center experienced a surge followed by a power outage which lasted several minutes from our electrical utility provider Progressive Energy. The surge tripped our facility’s main breaker; this main breaker is designed to have a certain level of sensitivity and to trip in the event of a severe surge in order to protect the load (servers and critical equipment) from being burned. Immediately after this occurred, our generator automatically started up within a few seconds. Meanwhile power to our load (servers and equipment) was automatically transitioned from unavailable raw power to generator power by the automatic transfer switch (ATS), our uninterruptible power system (UPS) in conjunction with our battery set supply is supposed to automatically sustain continuous power to the load. However, it appeared this did not happen. In any case, generator power was indeed immediately available within the minute of the outage.

Immediately post the outage our engineers and electricians came on site. The diagnosis conducted revealed there was a fault within a battery string which is connected to the UPS. It is this fault that disabled the UPS from being able to fully sustain continuous power to the load meanwhile the ATS transitioned the facility to the generator power lines from the raw power lines. During this time a great portion of the data center experienced a sudden power loss which caused a myriad of servers to power cycle. Unfortunately, at times when some systems experience sudden power loss some require manual administrator intervention to get full function restored. Post the outage, our team immediately started working on checking systems and all servers that may have been adversely affected by the sudden power loss these experienced.

DimeNOC Orlando from HostDime.com on Vimeo.

What was done to correct the problem:

Our on call UPS maintenance technician along with our electricians and engineers immediately came together on site to conduct a thorough diagnosis and put together a plan of action to correct any and all possible issues.

While the age of the battery supply being employed was well within the manufacturer’s life span expectancy, the entire battery supply was replaced with a new set. In addition, our UPS underwent a thorough in depth inspection and all critical components were individually inspected and reconditioned as necessary. Lastly, the batteries and UPS were load tested before being re-employed to the overall power back up system to ensure 100% reliability. All this was completed within several hours of the incident.

Who was affected:

The power outage experienced was intermittent. However, once power was fully restored to the facility many servers required file system checks (FSCK), some power supply replacements, and a few others hard drive replacements due to excessive I/O errors. Unfortunately, depending on the space on the drive the system occupies a FSCK run time can range from 30 minutes to a nine hours plus (approximately 200 servers counted). Those that were worst affected are the systems that were having excessive I/O errors and needed hard drive replacements (approximately 12 servers total counted). Again, unfortunately, hard drive replacements may take 4-12 hours plus to complete depending on the space being occupied on the drive. Those that were least affected were servers that only required a power supply replacement (approximately 60 servers counted).

For those servers that experienced the greatest downtime was not due directly to power unavailability, but rather due to post sudden power loss adverse effects described above.

What preventative measures are being taken:

All critical power systems in our data center and loads were previously and are regularly inspected and maintained. This includes generator, UPS, breakers, etc. In fact, our UPS underwent an inspection and a maintenance service on the week of the 12th of May 2008. The service report came back showing the UPS was in good working condition as well as the battery supply set. The only advice made was to consider replacement of the battery set supply as these were approaching the last year of the manufacturer’s life span expectancy. Pro actively following up on the advice made by the maintenance engineer, a new battery supply set was ordered right away and scheduled to be installed this Tuesday May 27, 2008.

Unfortunately, the battery supply set is what ended up being the fault and ironically this is what was already schedule for routine replacement maintenance. It is difficult to state that more could have been done as the batteries were within their life expectancy limits but failed short during this situation. Something of this magnitude, unfortunately, could not be predicted and was already being addressed with a new battery supply set replacement as a proactive measure. Nonetheless, a new standard has now been adopted as we will be increasing the battery reliability tests schedule to be completed monthly. This will allow us to intercept any and all types of possible issues with any battery sooner and overall highly reducing the probability of a failure encounter during critical times.

Our data center employs a 500KVA UPS and a 500KW generator. This is a statement that can be further proved by the recent pictures and videos taken yesterday afternoon. If you are in any kind of doubt whatsoever with regards to this, we would like to kindly ask for the opportunity to disprove your doubt. The pictures and videos below are of our backup systems in place which have protected us from several past outages to the entire data center. We uncover what maybe some of you didn’t know was in place in our facility since day one so you can see that your services with us are secure.

We have been in the industry close to 8 years now and we have always tried our best to ensure 100% uptime to all of you. This is the first outage we experienced with this level of severity in our entire existence. It is not only our job but our passion to give you the best level of service possible. We do not want to use the misfortune of this unpredictable situation to be an excuse for the downtime experienced. Despite the nature of the situation, we accept full responsibility for the outage and we are ready to compensate you in anyway we can. We value your business relationship and the level of trust you put in us. We know many of you will have a desire to cancel with us due to the losses you have incurred and question our systems’ integrity. We ask you to please talk to someone in management before you make your decision as we do understand the level of importance this means to each one of you. We work in high a high volatile environment where anything can happen just like with any of our competitors, however, we will always, no matter what, promise to be here whenever any issue occurs with an open hand to help resolve it as fast as humanly possible. Misfortunes will always happen to the best of us, how they are handled and treated makes the difference. If there is anything at all we can do to help you minimize your losses please just ask and consider it done. Our awareness and commitment level has tripled as a company and you can ensure this has only made us stronger and more experienced as a company. It is not everyday people or companies can overcome such issues and have the support and loyalty that many of you have given us. If you wish to reach out to me personally with any concerns, recommendations, suggestions, venting, or ways we can compensate you, please email me personally at e.v @ hostdime.com. I will be happy to talk to you in person.

__________________
Emmanuel, CTO
Surpass Hosting

While annoying at the time of failure, I truly admire a company that not only accepts responsibility for what happens, but doesn’t point the finger at the suppliers, contractors and everyone else. Way to go HostDime/Surpass!

It all started innocently enough. I was having some electrical work done that required a scheduled power shutdown for a few hours. While I do have an Uninterruptible Power Supply (UPS) for my home network, because the UPS run time is ~30 minutes for my network (well it is a home network after all) I prepared my network for a full shut down starting with my file/media/backup server, PCs then finally the network gear. All devices slowly took a final gasp and fell dormant as expected.

Once all the work was completed I reversed the process starting with the network gear, then the server, then the PCs. The network gear immediately sprang to life with that amber and green glow that makes all networking Geeks start breathing again. That’s were the warm and fuzz stuff ended. When I reached my server I pressed the power button. The green light came on…and that was it. Nothing else. No fan whirling. No flashing DVD/CD access lights. No HDD disk activity indicators. No grinding drives. Nada, nothing, zilch. Thinking that as with any elevator if I pushed the button again repeatedly in rapid fire succession that my data warehouse would spring to life, I pressed it again…and again…and again. There it was, my faithful server, staring at me and completely lifeless.

Now I had a few choices to make to ensure that I can access and use my data as soon as possible while expending the least amount of time and funds.

I could:

1. Begin a trial and error diagnosis
2. Restore Gigs of date from DVD media
3. Restore Gigs of data from Mozy’s off site servers
4. Remove the server’s hard drives and copy data from them using an external enclosure

As I was contemplating the above options and formulating a few others, I remembered that I had enabled network file sync on my Vista Desktop box to gauge its usefulness, performance and all around function. So, I decided to give it a try. I opened Explorer and navigated to a network share on my now dead server. Sure enough, Vista returned the folder and file structure that I was use too. I was able to open files and folders as if the server was still running. But these were locally cached versions. I then remapped the My Documents, My Pictures, My Music and other local pointers back to their original locations and simply copied the cached server files to the appropriate desktop locations. For those that I could find or that seemed out-of-date, I used Mozy’s handy restore option and slurped them from the Mozy server. An hour or so later all my data was back and accessible.

How does this make me greener you ask? This whole process had me reviewing my need for a server all together. So, I went out and grabbed a WD Caviar® GP 500GB HDD for document and media storage with appropriate share permissions on my desktop instead of a server. I pulled the 320GB and 160GB hard drives out of the now dead server. I’m using the 320GB in the newly purchased enclosure as a backup drive. The 160GB is currently offline awaiting a suitable use. I now have my desktop run Mozy for offsite backups instead of the server running it.

Will I be serverless forever? Only time will tell. However, after a couple of weeks the final result is one less computer running and using energy and I have my data where I need it as before. So far it looks like a win-win to me!

Some of my clients are not for profit organizations. These clients often have unique and specialized needs and often operate on shoestring budgets and with very small staffs.

As a contracted IT consultant, these organizations often are looking for me to provide solid, easy to use and cheap solutions. Recently I have begun in earnest recommending Google Apps for email, calendaring, collaboration and communication. The Google Apps suite offers easy tools for free. For non profit organizations, Google even offers a free upgrade to the Education Edition for those organizations that qualify.

Here is a very short overview of the steps that I usually take to get my clients up and running in no time:

1. Register the organization’s domain at an ICANN certified registrar (if they don’t already have one)
2. Register the organization for a free Google Apps Education Edition account
3. Setup email accounts within Google Apps and distribute usernames and passwords to organization’s staff
4. Create/edit domain MX records for Google Apps
5. After DNS propagation, test the administrator email account
6. Notify staff the they may begin using their new accounts

I’ve been using Google Apps for nearly a year. In that time I have had no client reports of missing mail, lost documents or issues logging into their accounts. So, while Google Apps may not meet everyone’s needs, It may just be the solution for smaller organizations that need enterprise level solutions at little or no cost.