Why you websites get taken down when under maintenance

Why do websites getting taken down when they are under maintenance? Surely, if you redesigning a website or something you can just redesign while keeping the old design up. Thanks

A properly setup up IT department maintains at least 3 environments

  1. DEV - all developement happens here… code is NEVER promoted from here to PRD
  2. UAT (User Acceptance Testing) … code is promoted from DEV. UAT mimics Production is all ways possible
  3. PRD (Production). Code is promoted from UAT once testing is completed

PRD is not offline unless the deployment process requires exclusive access or if server needs reboot… but this minimizes negative impact on clients. Some times there is a 4th [BUG] environment so that NEW code is created in DEV, while BUG is used to fix the current version.

Then add another level of Dev called Lab (For Network Engineers) which is a true sandbox for engineers. I hate when I have engineers destroy my dev environment thinking that it is a lab. :slight_smile:

One of the reasons is if the maintenance is further upstream. For instance, when our credit card provider is doing maintenance, they will often tell us that there may be intermittent outages during the maintenance period. We will automatically close the store during these periods to make sure no one has a bad purchasing experience.

Also, if the content of a website is backed by a database and the maintenance being done is on the database itself (like a software upgrade or a massive product-line update), it’s often easier to take the site offline for a little while, make all of the updates/changes, and then bring it back online all at once. Apple does this when they roll out most new product lines.

When I was a DBA at Fidelity Investments we had four sites with replicated databases. On maintenace nights we would route all traffic away from one of the target replicated sites and make all of our database/application software changes there and then test for 4-5 hours. If everything looked ok, we would route all traffic to the updated site and then update all of the other sites. We had no downtime.

Unfortunately, most companies can not afford to have four replicated sites, so they take their sites down and do the updates and then test.

When you have 20 million customers and 4.2 trillion dollars under management, you can do some really cool stuff.

And sometimes just not necessary. Fidelity has to serve customers 24/7 because financial instruments are traded all over the world 24/7. There are penalties if a securities transaction can’t be completed or a book can’t be closed. Companies selling merchandise don’t want to be down any longer than needed, but there are no penalties for not being able to sell some widget.

No. Fidelity Investments is open for business during the U.S. stock exchange hours all other business happens off-line. If the stock exchange were to drop 700 points (the amount might have changed since I left), Fidelity stops business for two hours and if the stock exchange continues to tank, it closes for the rest of the day.

It’s important to note that just because you purchase a fund (group of stocks) doesn’t mean that a buy is actually executed, even though you are charged for a buy (or sell). If Fidelity (and I’m guessing all companies like Fidelity) has sufficient funds(group of stocks) available, they just transfer it to you, charge you for the trade (or buy). These things happen in realtime, but if they need to get funds, it happens during normal business hours.

The cool thing about Fidelity Investments is that it ran more like a technology company than a financial company. They were deeply committed to the current trends in technology.

As an account holder, I think not. I’ve gone online to make funds transfers at all hours. Fidelity also supports after hours trading although I’ve never done that. I agree that security exchanges are only made during the hours the exchange is open, but customers can place orders at any time. Also, funds transfers are only executed when the Fed is open, but customers can place the orders anytime. I’d be moving my account if that were not possible.

If the market tanks by a large amount, it is the exchange that stops trading, not Fidelity. Fidelity does not have control over this.

I agree that Fidelity, like all large financial institutions could be confused with technology companies. I worked for a major NYC bank during my previous life and technology drove the business. Our users did not want to bother with any new business that could not be automated. At the time, we didn’t do business on the web and we did have to take our systems down everyday for batch processing. It was something the users were willing to pay big money to minimize. I’m sure by now they’ve solved that problem without me. :slight_smile:

Entering an order & having it executed aren’t the same thing though.
I can enter orders any time of day as well but they won’t be executed tip the next open business day when the exchanges open.
That may be peculiar to Canada though.

I have worked in an environment that had to be up 24/7/365 and doing maintenance in that set up was seriously complex as you had to make sure everything in the control center would fail over to the back up by doing a start up of the backup center.
Once you could prove that it could some up & take control then you failed everything there by redirecting traffic (fortunately you co do that in a very short amount of time) the backup ran operations and then when you were done you had to restart the control center & prove that IT came up OK before failing back.
Simple maintenance could take a week to perform because of all the checks you had to do on either end before failing operations back & forth.

But when you move a few billion cf of gas it’s important - otherwise stuff goes boom

I agree, but the website must (and does) remain up 24/7 to take orders. For a company like Fidelity, they may not be trading on foreign exchanges - I don’t know personally - but many brokers have their sites up to do transactions all over the world on exchanges they deal with.

The point is, it’s the kind of business you are doing that determines whether or not you remain up 24/7. Cost is secondary. If you can get by without needing it, you don’t pay the cost. If your customers demand it, you pay or go out of business.

[quote=50396:@Steve Albin]
The point is, it’s the kind of business you are doing that determines whether or not you remain up 24/7. Cost is secondary. If you can get by without needing it, you don’t pay the cost.[/quote]
In our case it was operations of a pipeline & yes - its just the cost of running the business.
Gas doesn’t stop moving at 5pm - and which 5Pm would that be anyways when you move it all across the continent ? :stuck_out_tongue:
So it was costly but it was just the cost of doing business

Or the govt says you have to do X which we have a lot of here :stuck_out_tongue:

[quote=44447:@Dave S]A properly setup up IT department maintains at least 3 environments

  1. DEV - all developement happens here… code is NEVER promoted from here to PRD
  2. UAT (User Acceptance Testing) … code is promoted from DEV. UAT mimics Production is all ways possible
  3. PRD (Production). Code is promoted from UAT once testing is completed

PRD is not offline unless the deployment process requires exclusive access or if server needs reboot… but this minimizes negative impact on clients. Some times there is a 4th [BUG] environment so that NEW code is created in DEV, while BUG is used to fix the current version.[/quote]

Most of the times promotion to Production is scheduled. Large insures companies I worked for had a 4 time a year schedule for promoting to Production. During a Promotion weekend ALL granted stuff was placed in production. So UAT had to be ready. Talking about careful planning. Missing the date resulted in a reschedule… 3 months to wait…

[quote=50399:@Alexander van der Linden]Most of the times promotion to Production is scheduled. Large insures companies I worked for had a 4 time a year schedule for promoting to Production. During a Promotion weekend ALL granted stuff was placed in production. So UAT had to be ready. Talking about careful planning. Missing the date resulted in a reschedule… 3 months to wait…

[/quote]
Interesting. Where I worked, it was much different. Production implementations were scheduled every week. The data center would limit the changes to go into production at one time because they wanted to limit the number or areas where there could be problems on Monday morning. Also, many times our production date was set in stone by a corresponding bank, organization, or regulator. No chance of waiting 3 months if things didn’t go well.