Re: getfsstat(2) MNT_NOWAIT & stale data for zpool

On 2025-Oct-08 15:59:12 +0000, Dave Cottlehuber <dch@skunkwerks.at> wrote:
>When does getfsstat(2) stale info get updated? It seems this can
>be easily more than 15 minutes, based on experiments in prod. It
>can be triggered by running `/bin/df` but not `/bin/df /`, which
>would be preferable.
>
>Background:
>
>I'm using prometheus node_exporter on a zpool to monitor disk usage.
>Right now this is broken, as a zpool can fill up, and node_exporter
>continues to return the old data.
>
>It loops over all filesystems, doing unix.Getfsstat(buf, unix.MNT_NOWAIT),
>which is a wrapper around getfsstat(2) which clearly states:
>
>     Normally mode should be specified as MNT_WAIT.  If mode is set to MNT_NOWAIT, getfsstat() will return the information it has
>     available without requesting an update from each file system.  Thus, some of the information will be out of date, but getfsstat()
>     will not block waiting for information from a file system that is unable to respond.  It will also skip any file system that is in
>     the process of being unmounted, even if the unmount would eventually fail.
>
>Is node_exporter just broken and should be fixed to use MNT_WAIT?
>
>Is there some tunable to refresh this stale data more frequently, 
>or should I just add `/bin/df` to a regular crontab?

See https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=273094 and https://github.com/prometheus/node_exporter/issues/1498

-- 
Peter Jeremy