Hi all,

It seems I managed to miss this bug report until Daniel copied me, apologies 
for that.

On 31/03/2026 13:37, Daniel Swarbrick wrote:
On Mon, 02 Feb 2026 18:42:47 +0000 [email protected] wrote:

 > after taking another pass at this issue, somebody pointed out that the 
Debian patches are indeed responsible for this behaviour: 
https://salsa.debian.org/go-team/packages/prometheus-node-exporter/-/blob/debian/sid/debian/patches/0001-Debian-defaults.patch?ref_type=heads
 >
 > These changes were migrated as part of report #908811. Now I wonder, are 
these overridden defaults still necessary today? The information scraped by this 
service should be cached by the kernel anyway, so waking up disks should not be 
the issue (anymore)?

Looping in Tina, since she was the author of that patch.

Yup, that was me many years ago, I first added that patch in 2018 :)

I guess the reason for excluding /media is that it typically has removable 
media dynamically mounted there, and usually you don't want such things to 
pollute your metrics. Using /mnt for permanently mounted filesystems is also a 
bit of an antipattern these days. At least in my experience, I would only ever 
mount something under /mnt temporarily, typically when trying to rescue a 
system or perform some data recovery (i.e. attaching a backup volume).

Yes, this was one of the main reasons, specially with things like docker 
mounts, per-user /run mounts, etc. that create lots of noise.
On top of that, I was also trying to avoid delays and timeouts caused by 
triggering auto-mounts or by accessing networked filesystems.


Of course, in the rare case that somebody is using /mnt for "production" 
filesystems, it is fairly trivial to override the default excluded mountpoints via 
command line args in /etc/default/prometheus-node-exporter.

I am undecided whether it was wise to move them to a source code patch, rather 
than ship them as a default option in ARGS.

I remember debating myself about this. I decided to patch instead for a better "user 
experience" out of the box, but it is not a very strong argument.


--
Martina Ferrari

Reply via email to