A rough account of what I did with Emacs recently.
- December 12
- Managed to complete a few backups eventually. So now I need to
figure out what I'm doing next; put the RAID unit back on the
older server, or move everything to the newer server? Maybe
this'll be a Christmas Holiday project!
The circus in the UK... they can't even properly organise a
revolt. It's all too stupid to make up. Oh, and BBC weren't
helping matters by trying to run "live" coverage of the outside
windows of the 1922 Committee's offices while the count was
happening, having the guy on camera saying inane things like,
"maybe their hands are cold and that's why it's taking so
long". Anyway, all over, now we're back to petulantance
vs. obstinancy while the slightly bemused, slightly confused rest
of the EU looks on wondering how the hell these people managed to
conquer a quarter of the planet only a few centuries
I have been noodling around with Advent of Code 2018, a fun
little site that gives you two programming puzzles each day and
builds up a picture as you solve the puzzles. I got stuck a little
a couple of days ago with one of the puzzles because I'd solved it
using brute force and then the followup puzzle was "now do the
same thing scaled up 100 times". After a couple of days it became
obvious that it wasn't going to complete any time soon, and worse
I'd not put any progress indicators in the code so I'd no idea if
it was at 9% or 90%. So I did a little hacking about this evening
(in the process discovering that Python allows you to do
array[index:] = new_array_values) and my canary run went
from 45 seconds to under 10, so I'm now rerunning it along with
some progress indicators and it's already 20% of the way
through. Shuold be done tomorrow, I guess. There's nothing
stopping me from moving on to the next puzzle in the mean time,
but I figured I'd do them in order.
- December 8
- Just doing a quick fsck...
/dev/rdisk2s2: fsck_hfs started at Sun Dec 2 10:54:05 2018
/dev/rdisk2s2: fsck_hfs completed at Sat Dec 8 07:12:15 2018
Now to see if it'll run a backup.
- December 7
- Company Christmas Party, where there was an actual old-fashioned
- December 2
- In an attempt to remedy the backup situation, I'm switching the
backup drive to a slightly beefier machine (for handwaving reasons
it needs to be a network share, not directly connected). That
means it's basically spent the last 11 hours verifying the drive,
and there's no progress bar to hint at when it might be
- December 1
- I have established a pattern of backup failure that goes like
So, the thing about this is that each time around the loop, the
probability of failure increases because there's more data to back
up (unless the machine you're backing up is entirely unused, and
even then it's still generating some level of churn on the disk
from time to time). I appear to have hit a tipping point on this
where I can't rely on the backup drive to be reliably available
for long enough to complete the backup.
- TimeMachine launches a backup
- Optionally, it spends many hours verifying the existing
backup. This might happen if the previous backup crapped out
- TimeMachine takes a snapshot of the drive to be backed up,
and estimates how much space it'll need on the backup
- TimeMachine discovers there's not enough free space, so it
starts pruning things from the backup drive, like the
previous failed backup.
- Time passes...
- TimeMachine starts copying files across, and things of that
nature (I will allow that TimeMachine does clever stuff
- Oh no! A spurious network event has interrupted the
connection to the backup drive! (good luck finding out what,
if anything, actually happened.)
- TimeMachine abandons the backup, leaving a .inProgress
backup occupying all that space it had cleared.
- Go back to the top.
I am so glad I have Arq as an offsite backup while I'm trying to
figure this mess out.
previous month | current month
Woop, 12-18 is here.