Our website is based around a customer based chat system. Customers are only on the system during a particular day and time range. So for parts of the week the server gets almost no traffic. When then time for a chat comes up we get hundreds of people on the site all doing page requests every few seconds.
So as you might imagine we are bringing the server to it's knees with heavy CPU and memory loads. Plus bandwidth usage is really high. We are currently on the biggest box that Rackspace has to offer. The site runs ok on it during these times. It's a little slow but not unbearable.
But we have not hit our max customer base. In fact if all goes well we will double our customer base next month. So I know when we do we will bring that box down totally.
I was thinking about possibly trying a cloud/cluster based approach but after some research on this site I have found that is probably not the best option.
So I am looking for advice on what to do? Is there a better host? Different technology?
Servers are not my thing really so I could really use some help.
No we can't change the way customers access the site or when they access. There is nothing that can be done in that regard so don't suggest it.
I am now with Rackspace's new cloud servers. I understand that they are still under a "beta testing" but I decided to give it a go.
Over all I like it. There are however some annoying parts which I will list first.
1. DNS manager issues... I have many DNS records for my domain and it seems that the DNS manager isn't so forgiving. It gives you the ability (or you think so) to delete an individual record however I have never been able to do that. It would always return with an error which is quite annoying. So I moved my DNS servers to another location and have been happy since.
2. Chat support is too much like dell tech support. The first person you get doesn't really know what he is talking about (I have been talking with them a few times about the above mentioned DNS issues and they confirmed multiple bugs within the system but this was at level 2 support and it was quite annoying having to explain the problem to everyone that I spoke too) What I ended up doing is going with Slicehosts chat room where I always found a very helpful hand (they are the same company so it's not stealing). They helped me with many of my issues.
3. The documents/wiki is very hard to find. If I didn't get a link to it from a tech support guy then I wouldn't of known it was there. All it is, is a wiki formatted silcehosts article repo. But it is quite helpful. My tip is to make it more visible.
Overall I am happy with the speed and stability of the server. So thanks Slicehost/Rackspace for your service.
im looking to add failover protection at my main server, in order that if it goes offline, all the traffic can be in a automatic way redirect to a failover / backup server.
For the failover server i was looking to buy a low-cost dedicated server located in other datacenter and have my dns be run at this failover server, but now i have been looking at Cloud Hosting, like Amazon or Mosso CloudServers and as it seems a very stable network, im thinking if it will be best to have setup/running my DNS at a Cloud Host...
So what do you think it would be best:
Option A: 1 main server 1 failover dedicated server 1 cloud (amazon/mosso) virtual server to run DNS
Option B: 1 main server 1 failover dedicated server (running the DNS)
I have been very interested in Mosso for quite some time, though Cloud Sites didn't seem quite right for what I needed with the compute cycles they had. However, their fairly new Cloud Sites looks very interesting, and their sales people at least will have me believe load balancing with several server instances will be superior to my current dedicated server.
Right now I have a server with Liquid Web that costs me $424/mo and 4x Intel(R) Core(TM)2 Quad CPU Q6600 @ 2.40GHz with 4GB of RAM. The average load on the server is anywhere from 30% at low times to 70-80% at peak times and memory usage is usually 20-50%. On average there are about 1000 mysql queries per second as the site is very ajax-intensive (hence Cloud Sites being way too expensive).
I don't really know the difference and technical side of all of this, I just program and do the business side of things, but I really like what Mosso has and am wondering if I would get a performance boost going with their Cloud Servers (Going with something like 8 server instances at 512mb RAM each @ only around $200/mo including bandwidth). Also, would I want to load balance all 8, or do something like 4 running the mysql and the other 4 serving the actual site?
obviously a (relatively successful) shared host would have multiple servers hosting their different clients. But they'd usually have all their client's setting their domain's nameservers to just one set of DNS (i.e. ns1/2.hostsdomain.tld)
Would you set this up via a DNS cluster? I.e. on a cPanel/WHM following these sorts of instructions?
For some reason I never can fully get my head around DNSes, so apologies if I ask any silly questions thusly.
With the above in mind, say your billing system sets up a new client - say it sets them up on SERVER2 (the host's second server for hosting clients). The DNS records would automatically be synced to the name server, right? And then, when someone visits the client's website/domain (with nameservers ns1/2.hostsdomain.tld) the 'hops' would roughly go:
Client's domain >> Host's name server >> This 'forwards' the 'connection'/request to the client's website files on SERVER2
Is that about right? And if the name server and SERVER2 are hosted in physically different parts of the US (say), would this make each request slower since it involves hops to two different servers before the client's files are shown?
I am planning for an architecture for the cluster of shared hosting servers. I intend to run around 10000 websites initially on Linux as well as Windows servers. I am planning for a load balanced setup and failover cluster with SAN or a similar storage solution where all the data will be stored. I am looking for suggestions on the best possible architecture which can accomodate all the above things. Also I need to know what are the best options available for load balancers and clusters.
This morning apache stopped to serve pages. I have tried to restart it with no success: ------- init.d/httpd restart ---------- init.d/httpd stop killall httpd init.d/httpd start ------ killall httpd init.d/httpd restart ---------
such like several times every time failed with the address already in use message
Was 4 oclock in the morning su I stopped investigating and restarted the machine
Is there a way to really shut down such a completely non-responsive process withou restarting the machine?
just a notice - there were not a heavy server load at the time
Server Version: Apache/2.2.6 (Mandriva Linux/PREFORK-8.2mdv2008.0) mod_ssl/2.2.6 OpenSSL/0.9.8e PHP/5.2.4 with Suhosin-Patch Server Built: Sep 12 2008 14:54:18
Code: ssh -fNg -L 8888:127.0.0.1:3306 email@example.com How can sort of "undo" that? I ask because I need to re-tunnel in to a different IP address but right now the only way I can figure to do it is to restart my computer (which is lame).
I have a question related DDOS attack. My hosting provider told me that my Server was DDos attacked few days ago. But in those days my server worked fine only apache server was down. The strange fact is that in the same day with this "DDOS attack" one of theyr admins worked something on SSL section of my server and during this operation the SSL hosts were down and httpd worked slow.
Inthe passed 3 months httpd worked very slow and after 2-3 restarts of httpd service the load droped down below 3.00 . I believe theyr httpd service was already with problems and that SSL configuration cause that apache failure in that day with "ddos attack"
I repeat in that day ONLY ssl hosts worked fine and non SSL hosts were down.
It's possibile on DDOS attack that load to be unde 0.5 , SSL hosts to work fine, FTP, Mail and other stuf to work like there is nobody on server (VERY FAST)?