When you stop a machine, we take great care to gracefully shut things down and ensure everything gets correctly written to your volumes. Very demure, very mindful.
But systems will crash, kernels will panic, and the filesystem on your volume can get very slightly corrupted. And when that goes unnoticed, it can cause trouble down the line.
That’s why we now check your volume for filesystem errors on every machine start.
You’ll need to update existing machines to take advantage of this, but on new machines you’ll something like this in your logs:
INFO Checking filesystem on /data
/dev/vdc: clean, 12/65280 files, 257024/261120 blocks
Occasionally, it’ll look a bit more like this:
/dev/vdc: recovering journal
/dev/vdc contains a file system with errors, check forced.
/dev/vdc: 12/65280 files (0.1% non-contiguous), 257024/261120 blocks
INFO Repaired filesystem on /data
This means we found some filesystem errors and repaired them automatically. All you have to do is enjoy the warm fuzzy feeling that some minor filesystem corruption won’t wreak havoc when you least expect it.