[pbs-devel] [PATCH 1/2] rrd: relay error to update database to caller

[Wolfgang Bumiller]

On Thu, Nov 14, 2024 at 02:29:49PM +0100, Thomas Lamprecht wrote:
> It does not make much sense to just log here, especially as the update
> fn has no context about what RRD series it's operating on.
> 
> I.e., logged message previously:
> > rrd update failed: time in past (...)
> 
> vs logged message now:
> > rrd::update_value 'host/cpu' failed - time in past (...)
> 
> The callers of the Database::update fn in the RRD Cache map can
> already handle errors, albeit it won't save the freshly loaded RRD in
> the map anymore if the update fails, any load will still do that
> though.

One of the call-chains comes from `apply_journal` (through
`apply_and_commit_journal_thread`). If the journal is corrupted
(negative time, NaNs) or the time is otherwise "messed up" (time in the
past).
Curiously the `journal_applied` result from that method 
return the *previous* time the thread ran. The thread will set this to
true if it succeeded.
So technically this could return "yes the journal is applied" while
logging "failed to apply the journal"... 🤔
This is a bit weird.
Maybe we should pass the name down the `apply_journal()` call chain so
it can log and skip like before?

> 
> Signed-off-by: Thomas Lamprecht <t.lamprecht at proxmox.com>
> ---
> It might be slightly nicer to factor out the common call to update to
> happen after getting/creating the RRD, but it's not trivial to do so as
> efficiently due to ownership handover when inserting the RRD in the map.
> 
>  proxmox-rrd/src/cache/rrd_map.rs |  4 ++--
>  proxmox-rrd/src/rrd.rs           | 14 ++++++--------
>  2 files changed, 8 insertions(+), 10 deletions(-)
> 
> diff --git a/proxmox-rrd/src/cache/rrd_map.rs b/proxmox-rrd/src/cache/rrd_map.rs
> index 0ef61cfa..4bcedade 100644
> --- a/proxmox-rrd/src/cache/rrd_map.rs
> +++ b/proxmox-rrd/src/cache/rrd_map.rs
> @@ -42,7 +42,7 @@ impl RRDMap {
>      ) -> Result<(), Error> {
>          if let Some(rrd) = self.map.get_mut(rel_path) {
>              if !new_only || time > rrd.last_update() {
> -                rrd.update(time, value);
> +                rrd.update(time, value)?;
>              }
>          } else {
>              let mut path = self.config.basedir.clone();
> @@ -61,7 +61,7 @@ impl RRDMap {
>              };
>  
>              if !new_only || time > rrd.last_update() {
> -                rrd.update(time, value);
> +                rrd.update(time, value)?;
>              }
>              self.map.insert(rel_path.to_string(), rrd);
>          }
> diff --git a/proxmox-rrd/src/rrd.rs b/proxmox-rrd/src/rrd.rs
> index 440abe06..4bf4f01b 100644
> --- a/proxmox-rrd/src/rrd.rs
> +++ b/proxmox-rrd/src/rrd.rs
> @@ -469,14 +469,10 @@ impl Database {
>      /// Update the value (in memory)
>      ///
>      /// Note: This does not call [Self::save].
> -    pub fn update(&mut self, time: f64, value: f64) {
> -        let value = match self.source.compute_new_value(time, value) {
> -            Ok(value) => value,
> -            Err(err) => {
> -                log::error!("rrd update failed: {}", err);
> -                return;
> -            }
> -        };
> +    pub fn update(&mut self, time: f64, value: f64) -> Result<(), Error> {
> +        let value = self
> +            .source
> +            .compute_new_value(time, value)?;
>  
>          let last_update = self.source.last_update;
>          self.source.last_update = time;
> @@ -485,6 +481,8 @@ impl Database {
>              rra.delete_old_slots(time, last_update);
>              rra.compute_new_value(time, last_update, value);
>          }
> +
> +        Ok(())
>      }
>  
>      /// Extract data from the archive
> -- 
> 2.39.5