All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
How to activate your site ASAP after the host dies?
In the past I avoided a situation where the host completely died because i had luckily shifted my websites few weeks prior to the meltdown.
However, this insecure feeling of abruptly loosing my vps (resulting in all the websites going offline) has always been nagging me ever since. Yes I do maintain backups but if the host dies then a new host would imply change in ips . The delay in setup of a new vps/propagation of ips/dns would lead to at least a downtime of 1 or 2 days.
The solutions I have read about range from establishing a vps proxy fetching data from 2 separate vps. There are ip fail over mechanisms & etc. But the thing is most of these requires costy setups, the other alternative is cloud like solutions from amazon and stuff but those are even far more expensive.
Sooo, what I want to know is;
**Is there anything that can be done to mitigate these problems ?
Can we speedily setup an alternate host with as little downtime as possible or
can we actually avoid any downtime by employing a cost effective solution to make sure our websites are always up and running without having to worry about server or host woes?**
Any answers would be much appreciated, I am sure the combined knowledge of all the hosting world veterans here will be of immense help.
Comments
Setup N mirrors and configure your DNS records to point to all of them?
by N mirrors do you mean multiple vps ? & how can we implement in the dns > by multiple A records ? In case of multiple A records does it not follow the first one it finds ?
Maybe.
Debatable.
Care to shed some light on your words of wisdom?
Yes
Client software should attempt to connect to the other records in case of one failing. Firefox does.
To implement failover would involve using multiple servers for redundancy, no? More servers means less chance of downtime, but that's all it is, a lower chance. Sure you could try and cover all the possible points of failure and have "no downtime at all" (aka run your own datacenter w/nuclear reactor and your own fiber lines), but at that point would it still be a cost effective solution?
Thanks MiguelQ. I vividly remember having tried this years ago without much success but from what you are saying maybe the browsers have caught on !
Still this does however pose the problem of browser compatibilities. I wish they made it proper part of the DNS protocol or is it already, otherwise why would multiple a records be allowed in the first place ? In any case, a solution then is to have multiple 'A' records with each record pointing to identical mirrors on different machines. But it requires at least 2 machines or more. I already have the dns setup on 2 different machines, I guess I should give this a try.
More alternatives?
haha so nicely put
then downtime is inevitable I guess 
But We could at least try to reduce it right to the best of our ability?
At that point doesn't it make more sense to go with a more reputable company? I fail to see value in attempting to save pennies by going with hosts that you may not fully trust yet, and then spending MORE in extra redundancy when you could just go with something you know won't dissapear overnight.
I completely agree with you, I'm currently with hostigation & inceptionhosting which I believe are pretty reliable hosts, still one should have contingencies in place for situations like a huge meteor striking the data center
but seriously bad things can happen & have known to happen.
I asked this same question a few weeks ago, as I am using my LEB VPS in a critical production environment. I am running my own asterisk pbx box which hosts 2 separate extensions. If my server goes down, so do the phones at 2 separate locations and I would not even know about it until someone called me on my mobile. That is WHY I asked the question. As modern as the internet is, and with it being built to withstand a Soviet nuclear attack, you would have thought that something as simple as this would have been catered for as the No.1 priority. But no one has a solution, other than monitor your server using another server and get it do send you an email if it goes down and use multiple DNS russian roulette. How sophisticated!
2+VPS
Round-Robin DNS
MySQL-Slave (or external MySQL-Provider with "close to real-time" Backup if you're lazy)
Just saying, I don't know why this seems to be such a big issue..
(yes, cheap solution and not perfect, but that's what was asked
)
thanks for the info and we have been pretty much discussing 2 machine dependent scenarios above
but maybe there is another way ? I have been thinking of dns ttls, do dns caching systems on isps rely on these ttls if the ttl values are too small ?
Could this be used to some benefit when it comes to dns propagation after quick expiry?
Use cloudflare and store files on an alternative host. Once your site drops just change CF DNS for pretty much instant results. CF also keeps it online if host goes down.
Use MySQL master-master.
2 (or more) VPS's that sync together + Rage4 DNS.
If you are hosting your website on LEB server then always take a daily backup to your local PC and If possible then purchase a another vps and add a extra nameserver to your domain so If your host dies then you can restore your backup to your 2nd vps/hosting account.
HTTP clients do not look for secondary A records once the first one is found. This only applies to NS/MX records.
To implement redundancy through DNS, you will need to setup two servers, obviously. Each server will host its own HTTP and DNS nameserver and point the A records it serves back to the same server (the primary nameserver will ignore the existence of the secondary nameserver).
If the HTTP server goes offline, so would the nameserver that resides along with it. So client machines will look for the 2nd nameserver that will obviously respond back with its own IP address on A records.
Your bigger problem will be synchronizing data between the two servers.
Let me know if you understood my rambling.
+
+
= Profit?
Cost of operating a real business.
I keep low TTL on my nameservers, and the following crons:
Fourth hourly backups of my databases, and rsync to a 500 GB storage server
Twelveth hourly file backups on the local server.
Daily rsyncs of the entire webserver and panel files, to the storage server.
File backups (1.7GB each) on the storage server are removed after every 15 days, but retaining 2 backups per month, just in case.
In addition, weekly backups are also transferred to a Swedish server.
Everything is run via scripts through cron jobs.
So far, it's been fine. However I think I should be better off splitting the 500 GB storage server into two seperate 250 GB locations.
Offline backups to my home PC are out of question considering that broadband is still costly around here. The 500 GB storage VPS costs around a fifth of the charges for a 10 GB 3G/broadband connection.
I never thought of it that way, it seems a lot more logical! two independent master-master dns servers instead of having to worry about browser compatibilities of multiple A records. Thank you, this I believe would do the trick
but I wonder what google and other search engines would think about this arrangement & is it possible that some systems think their might be an error with both dns values since all values are mostly thought to be copies of ns1 ?
wow talk about being meticulous. Awesome backup arrangement. This does however mean plenty of servers and reasonable I/O requirements. PM sent asking for further advice on server providers (Do mention your providers here instead if you think this should be public knowledge). Anyways, thanks for the comment, I am going to implement a similar arrangement depending on the resources I manage to acquire.
I was curious since I could follow two routes
i) Multiple A records with identical dns records on ns1 & ns2
or
ii) Single A record per ns but different IP on each ns servers
Interestingly both would work but my fear with the second one is the risk of the dns being identified as an IP mismatch since in most cases the A records are the same on all nservers.
So, I did a bit of digging and checked what everyone else was doing. Its Multiple A records with identical DNS except for google's ipv6(dont know why) , ipv4s are same all around.
**SO Option 1 IT IS **
@InfiniteTech Simply not true. Modern web browsers will try a second IP if the first IP is down.
Feel free to check out for further evidence...
chrome://net-internals/#dns
didnt worked for me either...
Use cloudflare to manage your Dns that way when you amend a record for the domain you are waiting 5-10 minutes for cloudflare systems to resolve. :P
Simples, DNS propergation by changing nameservers each time putting dns on same as website server caused 24-48 hours propergation time if user visits ip before you issue change as there dns server caches it.
personally id recommend using cloudflare to manage dns, you can even just click disable cloudflare, So you purely use it for DNS management an not the cloudflare frontend protection.
Taronyu this is true, Here is example of Own domain.
host ip class ttl
vpscorner.co.uk 141.101.127.182 IN 300
vpscorner.co.uk 108.162.200.183 IN 300
Two arecords for same hostname, if first is down then it defaults to second.
The same effect can be obtained, if you keep your TTL low on your own nameservers. That's what i do. I keep it at 3 minutes. The server seems to handle the extra load fine.
yes if you load your own DNS servers externally, Loading em on same site as host means its down while you move things over, I personally use cloudflare as it keeps it simple an easy.
I thought nobody does that - keeping DNS service on the same server. That's just stupid, and useless. I keep 5 nameservers, each one hosted on a different provider/DC.