My Raspberry Pi based hosting came to an abrupt end earlier this week as the RPi3 suddenly became unresponsive. Powering off and on the device resulted in an infinitive loop of I/O error messages. I’ve tried to recover the filesystem, but unfortunately my attempts proved to be unsuccessful.
I recently bought a new Raspberry Pi 3 and installed Slackware ARM current (hard float) on it. My goal was to compare the performance of the hard float port against Slackware ARM 14.2 (soft float), which is currently powering this RPi3 hosted website.
This website recently celebrated its second year of Raspberry Pi based hosting. It’s currently running on a RPi3 with Slackware ARM 14.2 (32-bit soft float). Somewhat to my surprise, the second year went by without a single glitch.
After four months of hosting this WordPress blog on a RPi3 I have yet to experience a single issue. No filesystem errors, no kernel oops’s and no hard freezes. Definitely an enjoyable change of pace from the preceding twelve months of hosting this site on a Raspberry Pi 2.
I’ve not been doing any overclocking whatsoever on this device and I believe that to be the decisive factor when it comes to improved stability. My old RPi2 is pretty much useless at this point and I’m guessing it simply got burned out before its time. As for storage, I’m still using a SanDisk MicroSDHC Ultra UHS-I 32GB card.
Some unfortunate but understandable news emerged on the Slackware ARM website today as ARM maintainer Stuart Winter released the following announcement:
Since I hit the one year mark today I thought I would do a quick update on my RPi2 project. A short recap to kick things off: the project had a rough start due to some overly ambitious overclocking that eventually resulted in severe data corruption. However, after implementing the necessary modifications I enjoyed close to 300 days of easy uptime before a power failure took the RPi2 down. My initial thought after the power failure was that everything was still dandy, but a few weeks later things started to go downhill fast. It all began with a few file system errors:
Due to a city wide power outage I lost just short of 300 days of uptime on the RPi2. The RPi2 did boot back up when the power returned, but since I had received a new IP address I needed to make a DNS update before the server was reachable again. That’s obviously the downside of running a server on a dynamic IP space, but hey it doesn’t cost me a cent. I have a 300 seconds TTL (Time To Live) on my blog.paranoidpenguin.net A record so I think it’s good enough for a hobby project.
Who would have thought, right? :-)
After installing my certificate from Let’s Encrypt last week I was immediately confronted with the fact that I had made the wrong choice in regard to key sizes. By using a 4096-bit private key I was relying too heavily on the RPi2’s CPU. This became abundantly clear as page load times were increased by 500 – 1000ms.
I woke up this morning to a mail informing me that WordPress had been upgraded to version 4.4.1. Shortly after I tried to access my blog to verify that everything had gone smoothly, but unfortunately my webserver showed no sign of life. Since I’ve previously had a few hard learned lessons with the RPi2, that made me a bit uneasy. A couple of hours later though, as I was reviewing my logs, the problem became pretty obvious:
So the last report from my Slackware based RPi2 hosting project ended on a cliffhanger (pun intended), as I was just recovering after suffering data corruption, the occasional kernel panic and random errors. Suspecting the instability might be caused by my overly optimistic approach to overclocking and overvolting, I decided to turn things down a few notches.
Excited by the prospect of hosting my blog on the new Raspberry Pi 2, I decided lately to wave goodbye to the local datacenter and unleash a Slackware Linux box into the wild (full story here).
Everything went (mostly) without a hitch until I wanted to get back in sync with the Slackware-current tree. After applying the available updates and issuing a reboot, the system seemed operational and nothing from the logs gave any indication of imminent failure.