I deployed a PG HA app with machines like a week ago and I’m 99% certain I reduced the memory from 4GB to 1 or 2GB (not sure now) with fly machines update.
But I checked and apparently they are still using 4GB?
I’ve reduced the memory again and will check in a couple of days.
Those are not in order, that’s a grouping of total usage for the current period ordered by size. You can see the current size of each machine with fly machine status <id>.
Only posted the image to show that the 2GB config only lasted a couple of hours, but the 4GB config lasted for weeks (even though I had changed it to 2GB right after creating the app).
I confirmed it today when doing fly machines update and the current config was 4GB. This is from today:
The first block in your screenshot looks like one app that had 4gb of RAM for a total of 44 minutes.
The second block looks like a second app that only ever had 4gb of RAM and ran for ~2 weeks.
Try running fly machine status <id> on one that seems wrong and check what the event log says. You should see update events when you resize Machines.
We actually have no code that would change a machine config automatically. Machines are entirely independent, it’s possible some bug could break one of them, but multiple machines getting updated together almost has to come from outside our infrastructure.
Initially I created a PG HA app and updated the config via fly machines update.
Then I deleted it because the default volume was too big (and volumes cannot be made smaller). So then created a new app with the exact same name but this time I forgot to update the config.
Of course the Usage section of the dashboard showed two apps with the same name. And had completely forgotten about what I did like two weeks ago.
Maybe in the Usage it would be a good idea to show something besides the user given app name (I guess you have like an internal id?).