[PATCH] metrics: remove experimental telemetry helper API
Stephen Hemminger
stephen at networkplumber.org
Wed Oct 7 17:32:10 CEST 2026
On Wed, 7 Oct 2026 09:03:23 +0100
Bruce Richardson <bruce.richardson at intel.com> wrote:
> On Tue, Oct 06, 2026 at 04:31:44PM -0700, Stephen Hemminger wrote:
> > While investigating experimental functions leftover in DPDK.
> >
> > The rte_metrics_tel_* functions are dead code. Nothing in the tree
> > calls them from outside the metrics library itself; they were left
> > exported by the telemetry library rewrite in 20.05.
> >
> > Stop exporting them. Delete the header, move the types into the C file
> > and make the helpers static. The file now only builds when jansson is
> > present, so the stub branch goes away too.
> >
> > The legacy commands ports_all_stat_values, global_stat_values and
> > ports_stats_values_by_name are unchanged.
> >
> > Signed-off-by: Stephen Hemminger <stephen at networkplumber.org>
> > ---
> > doc/guides/rel_notes/release_26_11.rst | 9 +++
> > lib/metrics/meson.build | 6 +-
> > lib/metrics/rte_metrics_telemetry.c | 106 ++++++-------------------
> > lib/metrics/rte_metrics_telemetry.h | 71 -----------------
> > 4 files changed, 39 insertions(+), 153 deletions(-)
> > delete mode 100644 lib/metrics/rte_metrics_telemetry.h
> >
>
> In the process, should we post a deprecation notice for the whole
> metrics-based legacy telemetry system? It's been superceded for many years
> now and it has a lot of limitations: supporting ethdev only, having
> globally registered stats (so if you have two different NIC types, you get
> a list of all stats names for both whatever port you query), and probably
> more beside. That would allow us to remove a large enough chunk of code in
> a future release.
>
> /Bruce
Right that was going as follow on patch, but can add it on to this.
More information about the dev
mailing list