Websites can go down at any time, and unplanned outages tend to happen at the worst possible times. But most professional Web teams try to make their planned outages as non-disruptive as possible. In fact, at many of the companies I've built websites for, even updating existing Web pages has to happen at off hours, especially if it's a major change. At one company I worked for, any change to the website that required the server to be stopped and restarted had to happen during our lowest traffic times. That was around 2am where I live. The idea was to minimize the impact. But another company I worked for has no problem taking parts of the website down for maintenance even during business hours. The overtime costs for developers working outside of business hours out-weighed the potential loss of revenue from taking the site down. What do you think? What times of the day is it okay to take a website or part of a website down for maintenance? If it depends, please explain in the comments.
This is a no-brainer … if you MUST take the site down then do it at the time of lowest (negative) impact.
However, it is not clear that one really has to take any site off-line for maintenance upgrades. If you create a mirror image of the ‘current’ site on your server (or host’s server) and change the DNS pointer to the new location – once it has ‘settled’ you can install & check out the maintenance changes on the primary location and then revert the DNS pointers back to the updated version when done.
You could also implement a site temporary redirect (in htaccess) but it would impeded checking out the maintenance version in its final location.
We’ve not had to do this – comments anyone?
I agree with Gerry.
If taking a site down is absolutely necessary (which should be quite rare) then do your homework and find the time of least visitor impact.
As an example, a customer of ours did need to take their site down in order to implement a new shopping cart. We had expected we would likely do this overnight but after reviewing their analytics data and talking with the site owner we discovered that the greatest sales from the site were occuring after business hours, generally after 6pm our time. This was a unique feature of his client base, they were hitting his site late at night with a peak from 6 to 11 pm West Coast time. We ended up taking the existing cart and sales pages down over the lunch period to give us ample time to implement, test and monitor.
I do like the idea of mirroring a site if that is possible for your situation but each situation will be unique. Often you can build new features or make substantial changes elsewhere on the server and do your testing without touching the live site until it is time to swap out. This can be done in minutes rather than hours. For something like a shopping cart, a 30 minute period is usually enough, staggering the paused elements to allow existing traffic to finish while preventing new instances and then re-implementing a pre-tested component.
I have never run into a situation where we would need to set up redirects on a temporary basis but I’m throwing that idea into the toolbox in case I ever do find a need.
Very helpfull!
Our servers and corporate offices are on the west coast and we are on the east coast. So, not sure of the planning they’ve done. Anyway they are moving to another data center, IP changes the works. Their plan is to start taking the servers down after midnight on a Friday and be back online by 1 pm the next day. To me this is an example of decision making in a vacuum. I just wonder how many big companies are like this? So really doesn’t matter what we as designers/developers and customer support think.
I would agree with Gerry that in the right situation a redirect would be great, however as Vancouver Web Design show there are many situations that this would not work.
Another example would be on just about any community driven site where the user is making changes ie.. posting to a forum. where a redirect would not be a good idea.
From a customer service point of view it would be best to do durring the lowest traffic times, but in some cases the cost of overtime may make it preferable to take a slight hit to your customer service rating.
The best time to take a site down is entirely dependent on the situation.
I think everybody here has valid points. I think, if nobody complained, then you have made a good choice.
I remember, once I took down a site (the PHP errors did, actually), that had mostly local audience, at around 3 AM local time, but somebody noticed and asked me about it.
I agree with Gerry. Its best to do this during non-peak times. Using google analytics or site statistics with your web host helps you to know when you have the least amount of traffic to your site. I would suggest making most changes on a test server or place on your hosting account. That way, the maintenance period is minimal because then you are only uploaded changed files, or new files. For our Atlanta Web Design clients, we try to employ these tools on there site. It helps with traffic and determining other things that are useful to their site and development.
As of a customer service point of view it would be greatest to do during the lowest traffic times, other than in some cases the cost of overtime may make it preferable to get a slight hit to your customer service rating.
Well, it depends on how long is the outage. A 5 mins outage can be acceptable almost all the time, however, a 5 hrs outage needs to be planned carefully, so it will not affect your business.
The idea is to take a website down when it encounters the least traffic, which is usually the time after midnight. But…. let me asked you (provocatively) – what is – the access time after midnight? Within the US the difference might be just 3 to 4 hours (depending on the Time Zone).
.
Yet, my own commercial website is accessed worldwide and potential customers come from the US while I moved my base recently from the US to Europe for a couple of years.
Here in Europe the time is 9 hours ahead of Pacific Time. So you do the math. No matter what I do, either Europe is effected or the US. Not to mention the time difference to Australia.
I know, all commercial sites are in the same boat but apparently nobody here thinks outside the US with their comments. Just food for thoughts
-hm
I really believe any answer other than “it depends” is almost insane.
“Least impact hours” seems a no-brainer, yes, but…
is this routine maintenance, or emergency maintenance?
If you have to take a web down as part of routine maintenance, something’s inherently wrong with your server setup. For example, my host provider uses a server cluster so, whenever a server is down for maintenance all the sites hosted on it are available through some other server within the cluster (and their dns services do their right thing to handle these cases). So, on the case of routine maintenance the best answer is “never”. If still you have to bring a site down for this then, yes, check your logs, see weekly/daily/hourly traffic ratios, and pick the lowest impact schedule.
For emergency maintenance, if things are serious enough to deserve taking the server down, they probably are serious enough for taking it down *immediately*. For example, if an e-commerce site’s SSL certificate becomes compromised, any secure form should be disabled immediately, and only re-enabled once a safe encryption key is up and working: even if this costs loss of sales, that’s still better than stolen client data.
In summary, the only answer that can be applied for all cases is “It depends”: server setup may interfere with your choice, and the reasons to take the site down may also be an important factor. If you have to take your site down, check all factors that apply to your situation before deciding.
Thanks for sharing your article. I really enjoyed it. I put a link to my site to here so other people can read it.