A Year with Endeavour

Reflecting on a year with bare metal

A little over a year ago, I rented a bare metal server from OVHcloud that I christened ‘Endeavour’ in this blog post. I thought I would offer some thoughts on the experience and things I would do differently, which has included deciding not to renew the lease on that server.

Good Things

It was honestly so cool to have that much power at my disposal. Plus the fact it was basically only mine. It was a whole server just for me, sitting in a rack in France somewhere. That was really nice to know. You can read the post I linked above to get a stronger picture of the specifications, but I definitely had more power in that one machine in comparison to any existing VMs I was running on cloud. If I was to get some kind of existing AWS VM size to match it, it would prove far more expensive than what the server was costing me per month.

It meant too, that I had room to grow. I was starting out this hobby where I was probably going to be spinning up quite a large footprint of tools, so I figured hey for the cost, it’ll be far easier to maintain the one machine than having multiple VMs that I would have to patch and keep an eye on. In this scenario of just me, having the one single point of failure would be an advantage!

Bad Things

Of course, nothing is perfect. Pretty quickly on, I ended up changing from using OVHcloud as my default public cloud provider, to other options. I found that the offering was just not mature enough and it felt a bit “janky” to use. That same “janky” feeling extended to the wrapper services if you could call them that, around the bare metal server. Specifically, the provided firewall services. For some reason there was a hard limit on the number of IP rules you could have which was extremely low. It was just hard to configure, though one could just rely on UFW, I wanted something to be stopping traffic before it even reached me.

Another thing was making mistakes in my configuration of the server hardware. Going with spinning hard drives so I would get the server provisioned faster, was a bad call since a lot of the tools I was running did not feel as snappy. Attempting to configure the 1Gbps private link that I purchased proved impossible to me with poor documentation not helping me either. Though I am willing to chalk that up to a lack of experience on my part. Availability was fine for the most part, I believe there was only two incidents where I lost contact with my server and both times were solved by a hard reboot on the OVH side.

Conclusion

In the end looking back, I definitely over provisioned and there was nothing I could do to correct that. In contrast to public cloud and VMs, if you’re oversized, you can easily resize yourself or the cloud provider will even recommend a size to you directly. I felt too that the money I was paying, roughly 140 EUR per month, if I was spending that on electricity, I could have far more local hardware which would rule out loss of service in the event of a fibre outage. My plans even included building out “Europa Server” however I ultimately put those on ice.

Soon though, I will be writing some more posts on my plans to bring everything I can in house. I’ve begun to lean more on Raspberry Pi’s for my needs and for x86 workloads I’m hoping to use some planned purchases to supplement the gap there. So, stay tuned to be able to find out more about what I have planned!