Accessing a Pi-Hole behind an Apache reverse proxy

Update 2019-09-15: Finally got around to looking into this and it turns out all I had to change was “ProxyPreserveHost Off” to “ProxyPreserveHost¬† On” to get things working. I’ve updated the original post to reflect the changes. I also didn’t note in my original host that I purposely restricted access to the apache virtual host to and (my internal networks). You’ll want to update the “Allow from” lines to reflect your internal networks OR remove the “<Location /></Location>” all together to make it accessible from anywhere (not recommended).

Update 2019-08-19: I just recently found out that this proxy configuration only allows read-only access to the Pi-Hole UI. I was attempting to white-list a domain and it was failing when accessing my Pi-Hole via the proxy. I had to go directly to the box’s FQDN to white-list a domain. I will leave this post for reference and update it when I figure out a fix to this problem.

Update 2019-09-29: My first Lets Encrypt certificate came due for auto-renewal and failed because of my original configuration. I’ve updated the apache configuration below so Lets Encrypt can access the non-SSL /.well-known directory to automatically renew certificates.


Original Post

Today I got tired of accessing my Pi-Hole over HTTP, having to remember to put /admin/ in the URL and having to load up a browse that wasn’t Vivaldi or Firefox because they don’t have an easy way to ignore Strict-Transport-Security for my domain.

I checked out some documentation about adding SSL to the Pi-Hole directly but have concerns that future updates will wipe out all the custom configuration to lighttpd. According to this you also have to be careful when enabling SSL on your Pi-Hole as it could interfere with blocking.

I already have an Apache webserver running so configuring it to reverse-proxy seemed like an easier task, plus if for some reason I wanted to access my Pi-Hole from the general internet (without VPN) it would be simple to enable that.

Here is the reverse proxy configuration I used with a restriction to my two internal networks and a redirect from HTTP to HTTPS:

I am aware that my SSL configuration is not the best. I’m waiting for CentOS 8 to come out before migrating off my existing CentOS 6 server.

To find the best SSL configuration for your OS and Web Server I recommend checking out Mozilla’s SSL Configuration Generator:

Some Microsoft Storage Spaces Benchmarks

My backup server has a ASRock H370M-ITX/AC motherboard in it and at the time of these benchmarks 3x6TB Seagate Ironwolf SATA disks.

I run Veeam and used the SATA disks as my backup repository.

My original configuration was 2x6TB Ironwolfs in a RAID0 (using Microsoft Storage Spaces) with Bitlocker enabled. This worked perfectly fine and I had no performance issues.

A project I was working on required me to add some redundancy to my backup storage so I purchased the 3rd disk and re-configured the Microsoft Storage Space as a RAID5 and re-enabled Bitlocker. Since then I had nothing but performance issues. When two backup jobs ran at the same time the server became nearly unresponsive. The jobs still ran and completed but it was very difficult to use the server and jobs took longer with the RAID5 configuration than the RAID0 configuration. A performance difference makes sense but this amount seemed abnormal.

I ended up with a spare LSI MegaRAID 9270-8i I couldn’t sell so I decided to throw it into my backup server and try running the above configuration with hardware RAID but before I did that, I ran some benchmarks.

As you can see BitLocker has a huge negative impact on this configuration even though the server is running a Intel Core i3-8100 which has hardware acceleration built in for encryption.

You can probably guess how this is all going to end now.

First a 3 disk RAID0:

Last but not least, a 3 disk RAID5:

It’ still a ~70% hit on sequential writes but the server is completely usable and backup jobs run at the speeds I would expect over 1GBe.