Hi,
Rather than burning tons of NVMe storage on my Mastodon cache I
thought I'd give bcachefs a go. It was upstreamed into the Linux 6.7
kernel in early January.
Seems to work.
$ uname -a
Linux social0 6.7-amd64 #1 SMP PREEMPT_DYNAMIC Debian 6.7.4-1~exp1 (2024-02-08) x86_64 GNU/Linux
$ sudo bcachefs format \
--label=ssd.ssd1 /dev/xvdd \
--label=hdd.hdd1 /dev/xvde \
--foreground_target=ssd \
--promote_target=ssd \
--background_target=hdd
$ sudo mount -t bcachefs /dev/xvdd:/dev/xvde /mnt/cache
$ sudo rsync -SHav /opt/mastodon/web/system/cache/ /mnt/cache/
$ sudo umount /mnt/cache
$ sudo mv /opt/mastodon/web/system/cache{,.old}
$ sudo mkdir /opt/mastodon/web/system/cache
$ sudo mount -t bcachefs /dev/xvdd:/dev/xvde /opt/mastodon/web/system/cache
$ df -h /opt/mastodon/web/system/cache
Filesystem Size Used Avail Use% Mounted on
/dev/xvdd:/dev/xvde 468G 61G 400G 14% /opt/mastodon/web/system/cache
xvdd is a 10GiB NVMe while xvde is a 500GiB HDD. It promotes stuff
to xvdd on read.
The kernel comes from Debian experimental. The only other things
that are needed are:
- about 140MiB of packaged dependencies
- a recent Rust compiler
- building bcachefs-tools from source (just "make; make install"
once you did the above two)
Of course, it may yet all end in tears.
No version of grub understands bcachefs yet so you wouldn't be able
to boot off of it.
https://bcachefs-docs.readthedocs.io/en/latest/index.html
Thanks,
Andy
--
https://bitfolk.com/ -- No-nonsense VPS hosting
Hi,
I've reconfigured our apt-cacher to use HTTPS for communications
with deb.debian.org since that is known to work well. So, all use of
deb.debian.org through our apt-cacher will use HTTPS between BitFolk
and deb.debian.org. You don't need to (and can't) change anything on
your end.
If you know of other in-use mirror sites that are reliable on HTTPS
then I'm happy to switch them over too. Reply here or send a support
ticket please.
More info: https://tools.bitfolk.com/wiki/Apt-cacher
Thanks,
Andy
--
https://bitfolk.com/ -- No-nonsense VPS hosting
Hi,
Since BitFolk moved datacentre from Telehouse to IP House, and since
some customers had previously actually asked about a renewable
energy statement, I've just updated it.
TL;DR: it's 59% now and they aim for it to be 100% by February 2025.
More detail:
https://tools.bitfolk.com/wiki/Renewable_energy_statement
Thanks,
Andy
--
https://bitfolk.com/ -- No-nonsense VPS hosting
Hi,
Between approximately 23:07Z and 23:14Z today, due to an error in
some work our colo provider was undertaking, two servers suffered
a total network outage and our remaining servers a partial outage.
Apologies for the disruption. It is not expected to re-occur.
The two worst affected servers were "clockwork" and "macallan".
Our servers have a pair of public network interfaces and are
connected to two separate switches. The error took out one of the
switches but left our ports enabled there, so it was a blackhole for
such traffic.
At the moment we use network bonding in active-backup mode. This
didn't fail over because the link state didn't go down, so the two
servers that had the misconfigured switch as their "active"
interface experienced the longer outage.
We are in the middle of transitioning away from a bonded setup and
having separate interfaces do BGP with our colo provider and use BGP
for such redundancy. We are already doing the BGP part but have yet
to disable the bonding and split the interfaces back out. When that
is complete — which I would hope to have done in a timescale of
weeks, not months — this failure mode won't exist.
Apologies again,
Andy
--
https://bitfolk.com/ -- No-nonsense VPS hosting
