Re: [pbs-devel] [PATCH proxmox-backup v5 5/8] api: add get_active_operations endpoint

[Dominik Csapak <d.csapak@proxmox.com>]

comments inline

On 1/24/22 13:31, Hannes Laimer wrote:
> Signed-off-by: Hannes Laimer <h.laimer@proxmox.com>
> ---
>   pbs-datastore/src/task_tracking.rs | 30 ++++++++++++++++++++++++++++
>   src/api2/admin/datastore.rs        | 32 +++++++++++++++++++++++++++++-
>   2 files changed, 61 insertions(+), 1 deletion(-)
> 
> diff --git a/pbs-datastore/src/task_tracking.rs b/pbs-datastore/src/task_tracking.rs
> index 06266e55..3f9693fa 100644
> --- a/pbs-datastore/src/task_tracking.rs
> +++ b/pbs-datastore/src/task_tracking.rs
> @@ -14,6 +14,36 @@ struct TaskOperations {
>       writing_operations: i64,
>   }
>   
> +pub fn get_active_operations(name: &str, operation: Operation) -> Result<i64, Error> {

imho it would make sense to get all types of operations back here,
that way we can save a call to this function later
(and with it the double parsing/locking (more on that below)/etc..)

> +    let path = PathBuf::from(format!("{}/{}", crate::ACTIVE_OPERATIONS_DIR, name));
> +    let lock_path = PathBuf::from(format!("{}/{}.lock", crate::ACTIVE_OPERATIONS_DIR, name));
> +
> +    let user = pbs_config::backup_user()?;
> +    let options = CreateOptions::new()
> +        .group(user.gid)
> +        .owner(user.uid)
> +        .perm(nix::sys::stat::Mode::from_bits_truncate(0o660));
> +
> +    let timeout = std::time::Duration::new(10, 0);
> +    open_file_locked(&lock_path, timeout, true, options.clone())?;

we only read and parse the file here, no updating, so we do not
have to lock here (the file is written atomically)

that way, the api call does not block a datastore operation

> +
> +    Ok(match file_read_optional_string(&path)? {
> +        Some(data) => {
> +            let active_tasks: Vec<TaskOperations> = serde_json::from_str(&data)?;
> +            active_tasks.iter()
> +                .filter(|task| procfs::check_process_running(task.pid as pid_t).is_some())
> +                .map(|task| {
> +                    match operation {
> +                        Operation::Read => task.reading_operations,
> +                        Operation::Write => task.writing_operations,
> +                    }
> +                })
> +                .sum()
> +        }
> +        None => 0,
> +    })
> +}
> +
>   pub fn update_active_operations(name: &str, operation: Operation, count: i64) -> Result<(), Error> {
>       let path = PathBuf::from(format!("{}/{}", crate::ACTIVE_OPERATIONS_DIR, name));
>       let lock_path = PathBuf::from(format!("{}/{}.lock", crate::ACTIVE_OPERATIONS_DIR, name));
> diff --git a/src/api2/admin/datastore.rs b/src/api2/admin/datastore.rs
> index ce710938..878f21d4 100644
> --- a/src/api2/admin/datastore.rs
> +++ b/src/api2/admin/datastore.rs
> @@ -43,7 +43,7 @@ use pbs_api_types::{ Authid, BackupContent, Counts, CryptMode,
>   use pbs_client::pxar::create_zip;
>   use pbs_datastore::{
>       check_backup_owner, DataStore, BackupDir, BackupGroup, StoreProgress, LocalChunkReader,
> -    CATALOG_NAME,
> +    CATALOG_NAME, task_tracking
>   };
>   use pbs_datastore::backup_info::BackupInfo;
>   use pbs_datastore::cached_chunk_reader::CachedChunkReader;
> @@ -1590,6 +1590,31 @@ pub fn get_rrd_stats(
>       )
>   }
>   
> +#[api(
> +    input: {
> +        properties: {
> +            store: {
> +                schema: DATASTORE_SCHEMA,
> +            },
> +        },
> +    },
> +    access: {
> +        permission: &Permission::Privilege(&["datastore", "{store}"], PRIV_DATASTORE_AUDIT, true),
> +    },
> +)]
> +/// Read datastore stats
> +pub fn get_active_operations(
> +    store: String,
> +    _param: Value,
> +) -> Result<Value, Error> {
> +    let reading = task_tracking::get_active_operations(&store, Operation::Read)?;
> +    let writing = task_tracking::get_active_operations(&store, Operation::Write)?;

this is what i meant above. when the function returns both operation counters,
we only have to call it once here

> +    Ok(json!({
> +        "read": reading,
> +        "write": writing
> +    }))
> +}
> +
>   #[api(
>       input: {
>           properties: {
> @@ -1947,6 +1972,11 @@ pub fn set_backup_owner(
>   
>   #[sortable]
>   const DATASTORE_INFO_SUBDIRS: SubdirMap = &[
> +    (
> +        "active-operations",
> +        &Router::new()
> +            .get(&API_METHOD_GET_ACTIVE_OPERATIONS)
> +    ),
>       (
>           "catalog",
>           &Router::new()