Sam Trelawny

Before you change the file locale, know what else it moves. Locale drives date parsing, the decimal mark and currency display, so flipping it on a file other people use can reformat dates that were entered under the old setting and quietly change what a number means. Fix your formulas, or make a copy and test the locale change there first.

Cost check: the smallest instances are around EUR 4, so under EUR 15 total you can afford a second small box for staging, or a managed database, rather than piling everything onto one machine. Splitting those was the best thing I did in my own migration.

Take off anything fully yellow and limp, since it is not feeding the plant any more and it rots against the stems. Leave anything with green left in it.

Sleeping thirty seconds after the work means your interval is thirty seconds plus however long the request took, so the schedule drifts. If you care about hitting a steady cadence, record the time before you start the cycle and sleep for whatever is left of the thirty seconds afterward, floored at zero. For a personal script the drift rarely matters; for anything logging on a timeline it matters a lot.

Whatever the number is, the contract does the real work. Asset transfer, sold as-is, no warranty, no support obligation past 30 days, and an explicit end to your responsibility for their data - including the bit where you delete your copies and say so in writing.

Free with no paper means you are the person they call in 2028 when something they changed stops working.

Measured our per-node tax so you have a shape to compare against. Logging agent about 300 MiB and a tenth of a core at our volume, node exporter tiny, kube-proxy tiny, CSI node plugin about 100 MiB, security agent about 250 MiB, service mesh sidecar injection is per pod rather than per node but adds up faster than any of these.

Call it half a core and just over a gigabyte per node before your workloads. On three 16 GiB nodes that is a bit over 3 GiB of overhead. On six 8 GiB nodes it is over 6 GiB for the same total memory. That is the real argument for consolidating.

Worth mentioning the registrar side is genuinely competitive right now: their .com list price sits around $11 and Cloudflare's at-cost is a bit under that. Both are about to move, because the .com wholesale fee goes from $10.26 to $10.97 on 1 November.

fair, and that matches my numbers being a modest improvement rather than none. it was worth doing, it just was not the whole story.

i did exactly this experiment. bundling and tree shaking took the artifact from 14mb to just under 4mb and cold start went from about 1.2s to 1.05s, which was real but not what i had been promised. the actual win came afterwards from moving two sdk clients out of module scope and requiring them lazily inside the handlers that needed them, which took another 300ms off because most requests never construct them at all.

Worth checking whether their once-a-month use is "I want to see it" or "I'd like to not think about it". If it's the second, a read-only shared view somewhere else is plenty.

The number nobody wants: you need per-category counts, not a total. 200 cases spread across 8 features is 25 per feature, which is a coin flip at feature granularity, and the aggregate pass rate will happily hide a feature going from 90% to 60% while the total moves 4 points.

Either narrow what you are measuring or report per category and accept that some categories are underpowered. One honest number per feature beats one dishonest number overall.

Six years, three different free hosts, one migration each time a terms change annoyed me. Total spend still zero. It's fine as long as the source lives somewhere you control.

For a 30-image portfolio you will hit none of them. Build minutes are the limit people actually reach, and only if they've wired up a rebuild on every commit for a site they edit constantly. 500 builds a month is an enormous number of deploys for a portfolio.