MikeN's questions

Running an Ubuntu server with MySQL for a high traffic production database server. Nothing else is running on the machine except the MySQL instance.

We store daily database backups on the DB server, is there any performance hit or reason why we should keep the hard disk relatively empty? If the disk is filled up to 86%+ with the database and all of the backups, does it hurt performance at all?

So would the DB server running with 86-90%+ full capacity perform less well in any way than the server running with only a 10% full disk?

The total disk size on the server is over 1 TB so even 10% of the disk should be enough for basic O/S swapping and such.

Does MySQL InnoDB engine use tablecache? I migrated from MyIsam to Innodb as the table engine and the monitors for "table-cache" hitrate are super low. Does Innodb use table-cache?

We are moving to two servers from one server due to performance issues. We are moving our MySQL DB to its own dedicated server and will keep the original machine as the front end machine running nginx/apache (Django backend.)

How do we monitor the value of this change? It is possible our whole site could actually get slower since the MySQL queries will be going out over a secure remote connection instead of locally?

www ~: free
             total       used       free     shared    buffers     cached
Mem:          2048       1848        199          0         21        274
-/+ buffers/cache:       1552        496
Swap:         4095       4086          9

On my Ubuntu production webserver that runs apache/nginx/MySQL and services lots of web customers, it looks like our free Swap memory space is almost all used up. What does this mean? I know that we should add more RAM/memory to the server, but I want to know how bad it will be if current usage continues? We were going to wait until this weekend, when Swap space runs out will the server crash? Grind to a halt? Run really slow?

I have a customer in Michigan who can't access my production SaaS webserver that is hosted on Slicehost. All other companies across the US/Canada/Europe have no problem reaching the site. This problem is occuring intermittantly, and Slicehost customer service says it's a problem with the client's ISP.

I got the IP address of my client, and ping'ing that IP address from my PROD server fails, but ping'ing the IP address from my dev box or our seperate blog server (also hosted on slicehost) works. How do I debug a problem like this? I asked the client to reach out to their local ISP and ask about this problem.

A traceroute shows that the packets are getting stopped on a Comcast Michigan node which is the client's ISP. Is there anything I can do additionally to fix this problem for my client?

I need to test sub-domains on my localhost. How can I effectively have this result of adding *.localhost.com to my /etc/hosts/ file?

If it's not possible, how do I work around this problem? I need to test wildcard sub-domains on my localserver. It is a Django devserver, can the Django dev server handle the sub-domains? Can some other piece of software/routing give me the end result I want?

I'm running a Django web application and have 4 instances of Apache running ModPython on my main webserver. I'm running Nginx on the front end and Nginx proxies all Django url's to the apache server pool.

How do I know if I need to run more instances of Apache to handle increased load? I've had a few complaints about timeouts and my first thought was that there weren't enough Apache instances running.

UPDATE: Correction, I'm running one instance of apache that has 4 child processes. They do show up as 5 distinct processes using "ps -ef | grep apache" on Linux.

How can I upgrade a really old Gutsy server to Karmic from the command line?

I'm seeing a very low percentage of misdelivered emails from my sendmail host. My sendmail server will try to email an address like "[email protected]" and the email will actually be delivered to "[email protected]" (which is my domain, the sendmail application is running on that webserver and is the email originator/sender domain.)

I've noticed this only happens for small and "weird" domains, which makes me guess that somehow the mail is undeliverable to the small domain because it's mail service is down, but just a total guess.

Has anyone seen this happen before? How can I diagnose and fix the problem?

I want to rewrite all http requests on my web server to be https requests, I started with the following:

server {
    listen      80;

    location / {
      rewrite     ^(.*)   https://mysite.com$1 permanent;
    }
...

One Problem is that this strips away any subdomain information (e.g., node1.mysite.com/folder), how could I rewrite the above to reroute everything to https and maintain the sub-domain?

Can you get an Extended Validation SSL certificate that applies to unlimited subdomains? Does it make technical sense? I'm using Digicert and they said that it wasn't possible to offer such a product.

I have some defined CNAME sub-domains on my site:
blog.site.com
wazoo.site.com
I want to add *.site.com support and have all of those wildcard requests get handled by my main web server (the sub-domains reflect other data, like the logged in user's custom data.) If you have a *.site CNAME, will that work well with the more specific records?
What I want is to have my blog.site.com CNAME work according to the DNS record and then have the wildcard of all the other user specific sub-domains go to my main website, is that possible?

How would I redirect all of the traffic from http://mysite.com/blog to a different server running at http://blog.mysite.com???

I want to do a 301 permanent redirect from http://blog.mysite.com to http://mysite.com/blog, but I want to keep the blog server running on http://blog.mysite.com, but I want it to appear to the user that they are viewing http://mysite.com/blog.

I just got a price quote from a VPS hosting company that told me my Virtual Private Server could host 10 simultaneous user connections at once. What would this correlate to in terms of burstable bandwidth (Megabits per second) that other companies quote? How could I convert between the two?

I could do a back of the envelope calculation:
1 page = 64 Kilobytes = 524288 bits
2 MBPS burstable bandwidth

So my effective simultaneous users per second with 2MBPS burstable bandwidth would be:

2E6 / 524288 =~ 3 users per second?

Is this a sane analysis? I'm assuming that once a connection with a client is established it is given priority until it is completed.

How can I calculate the required bandwith and cost of hosting a site that would have around 3 million unique visitors per year? I wanted to come up with an estimate of how much cost could be saved by using a cloud computing service like Amazon EC3 vs. using a traditional colocated server.

I'm serving up most of my static media from a CDN for my website. This includes images, videos, javascript files, and CSS files. YSlow and Google's Performance tool tell me that I should be gzip'ing this content. How can I gzip content served up by a CDN? Is there a way to store the content zipped on the CDN?

Would gzip'ing content from a CDN improve page load times? I don't care about the CDN bandwidth, but would be interested if it could actually load the page faster for the end user.

I have an Ubuntu 7.10 slice on Slicehost running my main web server. I want to update to Ubuntu 8.10 because the repositories are gone for 7.10 and I want the latest security and application upgrades.

Slicehost does not seem to let me dist-upgrade from 7.10 to 8.10, they suggested I buy a new VPS slice, upgrade it to what I need, port my data over, and then reimage my original slice from the 2nd slice. Has anyone else done this? Any better way to do it or pitfalls to watch out for while doing this? How long will my site be down if I use this method? Can I finish it in 2 hours on a Saturday?