Lost 72 hours worth of research data to forced update | Hacker News Reader

Lost 72 hours worth of research data to forced update

old.reddit.com

31 pointsest3119 comments

I'd say this person is actually lucky Windows forced an update. Compute nodes fail, which is why you log state so you can restore your simulation or training routine. Instead of experiencing a failure in industry, they learned the "cache state" lesson in a low-stakes academic setting. Maybe they'll extrapolate the lesson and start keeping off-site backups too.

Well this is not applicable to all scenarios/available hardware. We don't even know what the computation was and how powerful a machine he had access to

But of course I agree, if you can, you should

If you have enough ram to keep the current state of your computation, you probably have enough disk to push it to, several times over (for obvious reasons, you should never overwrite your latest snapshot with the most current data). If your computation was on a self-managed windows version, you definitely have enough disk.

Flash back to a few days ago when my Macbook rebooted and started the upgrade/patch process soon after a Zoom meeting ended. Lost 30min of work then went into a boot loop

Suprised me - I didn't know MacOS did that. Happened to my boss as well

That behavior is easy enough to turn off at least.

I have the opposite problem with macOS. It tells me there's an update, I tell it to do the update later expecting it to take care of that overnight, come back the next day to another notification telling me the update wasn't applied.

Assuming a process won't fail or a machine won't reboot for 3+ days is a bad idea on any OS, although it's certainly a worse idea on one that's targeted at the general public who only use Chrome and Office.

Writing the code in a way that can't be resumed is the person's second mistake.

Given this anecdote, I still prefer auto-updating Windows because it keeps everyone more safe. It's terrifying that there are millions of machines running Windows with access to my personal financial and medical data, and many of those may have 0-days on them. I don't know how we got by when those devices sometimes weren't fully updated for yeads.

I just want an update flowchart. Show me all the timers, all the conditions, all the paths, with a little marker that says "you are here".

I still haven't figured out which (if any) group policy settings will prevent automatic reboots. Hours of my life wasted.

at least in my experience, setting it through group policy so that it will not download patches without user interaction works. without downloaded patches it also won't install any automatically.

You need to tell IT to use WSUS. I've managed to convince IT to stick one my test stations into a 'never update' bucket. Months and months of uptime! :)

And the most vulnerable device around...

Its not, but thanks for your concern.

It's not exactly hard to prevent windows 10 from running updates. My security cam server has been running for years with updates only applying when I manually run them.

I do wish that Microsoft made it easier to prevent auto rebooting updates, but you can do it and it's pretty stupid to run a multi-day process without bothering to set up the system correctly and then bitch about it.

Always make long processes resumable. If it wasn't Windows updating it could have been a power failure or the process or OS crashing.

Surely there must be a way to chunk the data and feed it into a restful, idempotemt and atomic service for batch processing.

Is dealing with updates like dealing with backups - everybody has to commit a mistake himself to take it seriously?

Anyone else lose work because they were forced to restart Firefox?

Automatic updates can be disabled.

I don't recall ever being forced to restart Firefox. It just puts a message in the menu that says "Restart to update Firefox". But I can let that sit there as long as I want.

EDIT: Actually just looked it up and I see how you could have the issue. If you're using your package manager to update instead of using Mozilla's updater, then it will perform the update without accounting for whether or not Firefox is running. That means that the next time Firefox loads something from disk, it will mismatch with what the running version expects.

Reply on news.ycombinator.com