| Planning | Testing | Setup | Wi-Fi | Top |
My Wi-Fi access point is self-built, not a commercial appliance. I have
been encountering a mysterious problem in which the underlying CPU locks up
with no forensic evidence in the log files. When this happened on my main
router, an Intel NUC6CAYH (Celeron J3455), I solved
it by migrating
to a Raspberry Pi 3B. But after a few years, the problem returned. Other
people also have the problem and attribute it to a bug
in kernel
memory management. (I'm inclined to judge it as that's not a bug, that's
a feature
.) See the Planning section for
more details. One person got relief by throwing memory at the problem: he
upgraded his Raspberry Pi 3B with 1Gb RAM to a model 4 with 4Gb. I've also
found that tha 3B is kind of slow in a desktop replacement role. So I decided
to make the same upgrade, except with 8Gb.
OLD STUFF: For computers I have two roles for audio-video playback, and one for light duty code development display, basically a thin client for my desktop. For these machines I want low power and easy maintenance, meaning that I want to use my existing distro, OpenSuSE, and existing system administration tools.
The Raspberry Pi seemed to be an ideal choice for these roles: unbeatable price, low power, and a lot of users who are getting good use from them, not only students learning programming and hardware integration, but for precisely the thin client and audio-video performance roles that I'm looking for. And OpenSuSE has exploded into supporting ARM processors; it has a normal distro repository for aarch64 (64bit ARM) including RPi-specific support packages, among other hardware.
However, it seemed that everything that was normal and simple on an Intel processor and motherboard, such as plugging in a USB hub, turned into a week or more of frustration. Although I got one of the playback nodes performing music reliably for a while, the other goals were not met, and when the music stopped, so did the Raspberry Pi's place on my net.
I'm fairly sure that there is no fault in the RPi hardware, and specifically, one of the drivers, probably graphics, is to blame. Likely in a year or two I can take the RPi's out of storage and I'll find that they perform much better. But until then they are relegated to the cold palace.
Update in 2020-01-30: The graphics issue is no longer seen, while vigorously testing a variety of web browsers. I'm returning one of the RPi's to the desktop role, under the name Holly.
Contents:
| Planning | Testing | Setup | Wi-Fi | Top |
