Macs crash after 49 days of uptime…
TL;DR by Jason Snell:
1 Like
I wouldn’t want to be the one that tells Marco about this.
A rack of 48 Mac minis now powers Overcast’s podcast transcripts, as developer Marco Arment bypasses cloud AI in favor of local Apple Silicon
2 Likes
A script that randomly reboots after 25-35 days would easily sort this out, especially when you build an infrastructure which is designed to deal with “failure”
2 Likes
I agree. As long as all needed systems can be started prior to login. You can’t automatically log in a user if FileVault is enabled.
1 Like
IIRC Marco said something about this now being possible via SSH?
3 Likes
At work we run one of the most reliable server platforms there is (no, it’s not any flavour of Unix). Only a couple of years ago did we move from monthly to 2-monthly maintenance windows (after 30+ years of monthly).
The general approach is to reboot (we do many other things in these windows, too) on a regular cadence before you know you need it. Even the most reliable systems can have unforeseen glitches or runaway processes.
Building a system that cannot be rebooted seems like a bad idea to me. I mean… I guess there will be some who thank them for tripping over this issue so no-one else has to.
2 Likes
Yes, that apparently is a new Tahoe feature (SSH over Ethernet only).
1 Like
It used to be a badge of honour “That Netware server has been up for more than 5 years without a reboot”
Then 2 things happened
- The internet allowed regular updates which brought both additional functionality and software/security patches. Before that updates arrived on Floppy disks/CDs/DVDs and were often ignored if nothing was wrong.
- The internet increased the threat vector for any server, not just those accessible via the internet.
5 Likes
What OS is used on the servers?
My guess is three letters followed by a lower case i
1 Like
I haven’t run into this on my home server, which is running Sequoia (It’s an Intel machine, and so can’t update to Tahoe) It only gets rebooted when there’s an update.