Re: [pbs-devel] [PATCH proxmox-backup 1/4] proxmox-backup-proxy: fix leftover references on datastore removal

[Thomas Lamprecht <t.lamprecht@proxmox.com>]

On 01.06.21 14:13, Dominik Csapak wrote:
> when we remove a datastore via api/cli, the proxy
> has sometimes leftover references to that datastore in its
> DATASTORE_MAP which includes an open filehandle on the
> '.lock' file
> 
> this prevents unmounting/exporting the datastore even after removal,
> only a reload/restart of the proxy did help
> 
> add a command to our command socket, which removes all non
> configured datastores from the map, dropping the open filehandle
> 
> Signed-off-by: Dominik Csapak <d.csapak@proxmox.com>
> ---
>  src/api2/config/datastore.rs    |  4 +++-
>  src/backup/datastore.rs         | 16 ++++++++++++++++
>  src/bin/proxmox-backup-proxy.rs | 11 +++++++++++
>  src/server.rs                   |  8 ++++++++
>  4 files changed, 38 insertions(+), 1 deletion(-)
> 
> diff --git a/src/api2/config/datastore.rs b/src/api2/config/datastore.rs
> index 6ca98b53..33e76b27 100644
> --- a/src/api2/config/datastore.rs
> +++ b/src/api2/config/datastore.rs
> @@ -403,7 +403,7 @@ pub fn update_datastore(
>      },
>  )]
>  /// Remove a datastore configuration.
> -pub fn delete_datastore(name: String, digest: Option<String>) -> Result<(), Error> {
> +pub async fn delete_datastore(name: String, digest: Option<String>) -> Result<(), Error> {
>  
>      let _lock = open_file_locked(datastore::DATASTORE_CFG_LOCKFILE, std::time::Duration::new(10, 0), true)?;
>  
> @@ -425,6 +425,8 @@ pub fn delete_datastore(name: String, digest: Option<String>) -> Result<(), Erro
>      let _ = jobstate::remove_state_file("prune", &name);
>      let _ = jobstate::remove_state_file("garbage_collection", &name);
>  
> +    crate::server::refresh_datastores().await?;

two methods named exactly the same, one async the other not is a bit weird, even
if in different modules. 

I'd, Either follow the name scheme from the cert command for the send command over
socket helper, e.g., refresh_proxy_datastores or notify_datastore_removed (in
theory the refresh of the internal datastore map may not be the only thing we
want to do (in the future)); just to name two examples for improving on the
duplicate function names

> +
>      Ok(())
>  }
>  
> diff --git a/src/backup/datastore.rs b/src/backup/datastore.rs
> index 584b2020..6989313d 100644
> --- a/src/backup/datastore.rs
> +++ b/src/backup/datastore.rs
> @@ -69,6 +69,22 @@ impl DataStore {
>          Ok(datastore)
>      }
>  
> +    /// removes all datastores that are not configured anymore
> +    pub fn refresh_datastores() -> Result<(), Error>{
> +        let (config, _digest) = datastore::config()?;
> +        let mut stores = HashSet::new();
> +        for (store, _) in config.sections {
> +            stores.insert(store);
> +        }

above feels a bit odd as the section config effectively is a hashmap already
and it's loaded anyway

> +
> +        let mut map = DATASTORE_MAP.lock().unwrap();
> +        // removes all elements that are not in the config
> +        map.retain(|key, _| {
> +            stores.contains(key)
> +        });

Single expression closures do not need { }

map.retain(|key| stores.contains(key)); // drop removed datastores

> +        Ok(())
> +    }
> +
>      fn open_with_path(store_name: &str, path: &Path, config: DataStoreConfig) -> Result<Self, Error> {
>          let chunk_store = ChunkStore::open(store_name, path)?;
>  
> diff --git a/src/bin/proxmox-backup-proxy.rs b/src/bin/proxmox-backup-proxy.rs
> index a53f554a..e2953417 100644
> --- a/src/bin/proxmox-backup-proxy.rs
> +++ b/src/bin/proxmox-backup-proxy.rs
> @@ -136,6 +136,17 @@ async fn run() -> Result<(), Error> {
>          },
>      )?;
>  
> +    // to remove references for not configured datastores
> +    commando_sock.register_command(
> +        "refresh-datastores".to_string(),
> +        |_value| {
> +            if let Err(err) = proxmox_backup::backup::DataStore::refresh_datastores() {
> +                log::error!("could not refresh datastores: {}", err);
> +            }
> +            Ok(Value::Null)
> +        }
> +    )?;
> +
>      let server = daemon::create_daemon(
>          ([0,0,0,0,0,0,0,0], 8007).into(),
>          move |listener, ready| {
> diff --git a/src/server.rs b/src/server.rs
> index ba25617d..aa068a07 100644
> --- a/src/server.rs
> +++ b/src/server.rs
> @@ -100,3 +100,11 @@ pub(crate) async fn reload_proxy_certificate() -> Result<(), Error> {
>          .await?;
>      Ok(())
>  }
> +
> +pub(crate) async fn refresh_datastores() -> Result<(), Error> {
> +    let proxy_pid = crate::server::read_pid(buildcfg::PROXMOX_BACKUP_PROXY_PID_FN)?;
> +    let sock = crate::server::ctrl_sock_from_pid(proxy_pid);
> +    let _: Value = crate::server::send_raw_command(sock, "{\"command\":\"refresh-datastores\"}\n")
> +        .await?;
> +    Ok(())
> +}
>