public inbox for pbs-devel@lists.proxmox.com
 help / color / mirror / Atom feed
* [pbs-devel] [PATCH proxmox-backup v2 1/2] api: avoid retrieving lsblk result twice
@ 2024-09-17  8:05 Gabriel Goller
  2024-09-17  8:05 ` [pbs-devel] [PATCH proxmox-backup v2 2/2] api: parallelize smartctl checks Gabriel Goller
  2024-11-12 20:18 ` [pbs-devel] applied: [PATCH proxmox-backup v2 1/2] api: avoid retrieving lsblk result twice Thomas Lamprecht
  0 siblings, 2 replies; 3+ messages in thread
From: Gabriel Goller @ 2024-09-17  8:05 UTC (permalink / raw)
  To: pbs-devel

Avoid running `lsblk` twice when executing the `list_disk`
endpoint/command. This and the various other small nits improve the
performance of the endpoint.

Does not really fix, but is related to: #4961.

Signed-off-by: Gabriel Goller <g.goller@proxmox.com>
---

v2:
 - nothing

 src/api2/node/disks/mod.rs    |  6 +++++-
 src/tools/disks/mod.rs        | 10 +++++++---
 src/tools/disks/smart.rs      | 10 +++-------
 src/tools/disks/zfs.rs        | 15 ++++++++-------
 src/tools/disks/zpool_list.rs |  2 +-
 5 files changed, 24 insertions(+), 19 deletions(-)

diff --git a/src/api2/node/disks/mod.rs b/src/api2/node/disks/mod.rs
index 4ef4ee2b8ac7..abcb8ee40ef1 100644
--- a/src/api2/node/disks/mod.rs
+++ b/src/api2/node/disks/mod.rs
@@ -115,7 +115,11 @@ pub fn smart_status(disk: String, healthonly: Option<bool>) -> Result<SmartData,
 
     let manager = DiskManage::new();
     let disk = manager.disk_by_name(&disk)?;
-    get_smart_data(&disk, healthonly)
+    if let Some(path) = disk.device_path() {
+        get_smart_data(path, healthonly)
+    } else {
+        bail!("disk {:?} has no node in /dev", disk.syspath());
+    }
 }
 
 #[api(
diff --git a/src/tools/disks/mod.rs b/src/tools/disks/mod.rs
index c729c26a20a9..04f62b818238 100644
--- a/src/tools/disks/mod.rs
+++ b/src/tools/disks/mod.rs
@@ -863,8 +863,8 @@ fn get_partitions_info(
     lvm_devices: &HashSet<u64>,
     zfs_devices: &HashSet<u64>,
     file_system_devices: &HashSet<u64>,
+    lsblk_infos: &[LsblkInfo],
 ) -> Vec<PartitionInfo> {
-    let lsblk_infos = get_lsblk_info().ok();
     partitions
         .values()
         .map(|disk| {
@@ -887,8 +887,8 @@ fn get_partitions_info(
 
             let mounted = disk.is_mounted().unwrap_or(false);
             let mut filesystem = None;
-            if let (Some(devpath), Some(infos)) = (devpath.as_ref(), lsblk_infos.as_ref()) {
-                for info in infos.iter().filter(|i| i.path.eq(devpath)) {
+            if let Some(devpath) = devpath.as_ref() {
+                for info in lsblk_infos.iter().filter(|i| i.path.eq(devpath)) {
                     used = match info.partition_type.as_deref() {
                         Some("21686148-6449-6e6f-744e-656564454649") => PartitionUsageType::BIOS,
                         Some("c12a7328-f81f-11d2-ba4b-00a0c93ec93b") => PartitionUsageType::EFI,
@@ -942,6 +942,7 @@ fn get_disks(
     // fixme: ceph journals/volumes
 
     let mut result = HashMap::new();
+    let mut device_paths = Vec::new();
 
     for item in proxmox_sys::fs::scan_subdir(libc::AT_FDCWD, "/sys/block", &BLOCKDEVICE_NAME_REGEX)?
     {
@@ -1009,6 +1010,8 @@ fn get_disks(
             .map(|p| p.to_owned())
             .map(|p| p.to_string_lossy().to_string());
 
+        device_paths.push((name.clone(), devpath.clone()));
+
         let wwn = disk.wwn().map(|s| s.to_string_lossy().into_owned());
 
         let partitions: Option<Vec<PartitionInfo>> = if include_partitions {
@@ -1018,6 +1021,7 @@ fn get_disks(
                     &lvm_devices,
                     &zfs_devices,
                     &file_system_devices,
+                    &lsblk_info,
                 ))
             })
         } else {
diff --git a/src/tools/disks/smart.rs b/src/tools/disks/smart.rs
index 3ad782b7b248..4868815f6f32 100644
--- a/src/tools/disks/smart.rs
+++ b/src/tools/disks/smart.rs
@@ -1,8 +1,8 @@
-use std::collections::{HashMap, HashSet};
 use std::sync::LazyLock;
+use std::{collections::{HashMap, HashSet}, path::Path};
 
 use ::serde::{Deserialize, Serialize};
-use anyhow::{bail, Error};
+use anyhow::Error;
 
 use proxmox_schema::api;
 
@@ -76,7 +76,7 @@ pub struct SmartData {
 }
 
 /// Read smartctl data for a disk (/dev/XXX).
-pub fn get_smart_data(disk: &super::Disk, health_only: bool) -> Result<SmartData, Error> {
+pub fn get_smart_data(disk_path: &Path, health_only: bool) -> Result<SmartData, Error> {
     const SMARTCTL_BIN_PATH: &str = "smartctl";
 
     let mut command = std::process::Command::new(SMARTCTL_BIN_PATH);
@@ -85,10 +85,6 @@ pub fn get_smart_data(disk: &super::Disk, health_only: bool) -> Result<SmartData
         command.args(["-A", "-j"]);
     }
 
-    let disk_path = match disk.device_path() {
-        Some(path) => path,
-        None => bail!("disk {:?} has no node in /dev", disk.syspath()),
-    };
     command.arg(disk_path);
 
     let output = proxmox_sys::command::run_command(
diff --git a/src/tools/disks/zfs.rs b/src/tools/disks/zfs.rs
index 2abb5176c1e6..3b7da1540835 100644
--- a/src/tools/disks/zfs.rs
+++ b/src/tools/disks/zfs.rs
@@ -70,7 +70,7 @@ pub fn zfs_pool_stats(pool: &OsStr) -> Result<Option<BlockDevStat>, Error> {
 ///
 /// The set is indexed by using the unix raw device number (dev_t is u64)
 pub fn zfs_devices(lsblk_info: &[LsblkInfo], pool: Option<String>) -> Result<HashSet<u64>, Error> {
-    let list = zpool_list(pool, true)?;
+    let list = zpool_list(pool.as_ref(), true)?;
 
     let mut device_set = HashSet::new();
     for entry in list {
@@ -79,12 +79,13 @@ pub fn zfs_devices(lsblk_info: &[LsblkInfo], pool: Option<String>) -> Result<Has
             device_set.insert(meta.rdev());
         }
     }
-
-    for info in lsblk_info.iter() {
-        if let Some(partition_type) = &info.partition_type {
-            if ZFS_UUIDS.contains(partition_type.as_str()) {
-                let meta = std::fs::metadata(&info.path)?;
-                device_set.insert(meta.rdev());
+    if pool.is_none() {
+        for info in lsblk_info.iter() {
+            if let Some(partition_type) = &info.partition_type {
+                if ZFS_UUIDS.contains(partition_type.as_str()) {
+                    let meta = std::fs::metadata(&info.path)?;
+                    device_set.insert(meta.rdev());
+                }
             }
         }
     }
diff --git a/src/tools/disks/zpool_list.rs b/src/tools/disks/zpool_list.rs
index 83ce063dfe16..485a721b9865 100644
--- a/src/tools/disks/zpool_list.rs
+++ b/src/tools/disks/zpool_list.rs
@@ -136,7 +136,7 @@ fn parse_zpool_list(i: &str) -> Result<Vec<ZFSPoolInfo>, Error> {
 ///
 /// Devices are only included when run with verbose flags
 /// set. Without, device lists are empty.
-pub fn zpool_list(pool: Option<String>, verbose: bool) -> Result<Vec<ZFSPoolInfo>, Error> {
+pub fn zpool_list(pool: Option<&String>, verbose: bool) -> Result<Vec<ZFSPoolInfo>, Error> {
     // Note: zpools list verbose output can include entries for 'special', 'cache' and 'logs'
     // and maybe other things.
 
-- 
2.39.5



_______________________________________________
pbs-devel mailing list
pbs-devel@lists.proxmox.com
https://lists.proxmox.com/cgi-bin/mailman/listinfo/pbs-devel


^ permalink raw reply	[flat|nested] 3+ messages in thread

* [pbs-devel] [PATCH proxmox-backup v2 2/2] api: parallelize smartctl checks
  2024-09-17  8:05 [pbs-devel] [PATCH proxmox-backup v2 1/2] api: avoid retrieving lsblk result twice Gabriel Goller
@ 2024-09-17  8:05 ` Gabriel Goller
  2024-11-12 20:18 ` [pbs-devel] applied: [PATCH proxmox-backup v2 1/2] api: avoid retrieving lsblk result twice Thomas Lamprecht
  1 sibling, 0 replies; 3+ messages in thread
From: Gabriel Goller @ 2024-09-17  8:05 UTC (permalink / raw)
  To: pbs-devel

To improve the performance of the smartctl checks, especially when a lot
of disks are used, parallelize the checks using the `ParallelHandler`.

Signed-off-by: Gabriel Goller <g.goller@proxmox.com>
---

Benchmark:
Benchmark 1: proxmox-backup-manager disk list (parallel)
  Time (mean ± σ):      75.8 ms ±   5.7 ms    [User: 0.8 ms, System: 0.5 ms]
  Range (min … max):    63.4 ms …  95.3 ms    100 runs

Benchmark 2: proxmox-backup-manager disk list (sequential)
  Time (mean ± σ):     189.4 ms ±   6.7 ms    [User: 0.7 ms, System: 1.0 ms]
  Range (min … max):   178.8 ms … 223.0 ms    100 runs

Summary
  'proxmox-backup-manager disk list (parallel)' ran
    2.50 ± 0.21 times faster than 'proxmox-backup-manager disk list (sequential)'

 src/tools/disks/mod.rs | 42 +++++++++++++++++++++++++++++-------------
 1 file changed, 29 insertions(+), 13 deletions(-)

v2, thanks @shannon:
 - fixed charset of email

diff --git a/src/tools/disks/mod.rs b/src/tools/disks/mod.rs
index 04f62b818238..09a3003a9333 100644
--- a/src/tools/disks/mod.rs
+++ b/src/tools/disks/mod.rs
@@ -15,13 +15,15 @@ use once_cell::sync::OnceCell;
 use ::serde::{Deserialize, Serialize};
 
 use proxmox_lang::{io_bail, io_format_err};
+use proxmox_log::info;
 use proxmox_schema::api;
 use proxmox_sys::linux::procfs::{mountinfo::Device, MountInfo};
 
 use pbs_api_types::{BLOCKDEVICE_DISK_AND_PARTITION_NAME_REGEX, BLOCKDEVICE_NAME_REGEX};
 
+use crate::tools::parallel_handler::ParallelHandler;
+
 mod zfs;
-use tracing::info;
 pub use zfs::*;
 mod zpool_status;
 pub use zpool_status::*;
@@ -1047,16 +1049,6 @@ fn get_disks(
             usage = DiskUsageType::DeviceMapper;
         }
 
-        let mut status = SmartStatus::Unknown;
-        let mut wearout = None;
-
-        if !no_smart {
-            if let Ok(smart) = get_smart_data(&disk, false) {
-                status = smart.status;
-                wearout = smart.wearout;
-            }
-        }
-
         let info = DiskUsageInfo {
             name: name.clone(),
             vendor,
@@ -1067,8 +1059,8 @@ fn get_disks(
             size,
             wwn,
             disk_type,
-            status,
-            wearout,
+            status: SmartStatus::Unknown,
+            wearout: None,
             used: usage,
             gpt: disk.has_gpt(),
             rpm: disk.ata_rotation_rate_rpm(),
@@ -1077,6 +1069,30 @@ fn get_disks(
         result.insert(name, info);
     }
 
+    if !no_smart {
+        let (tx, rx) = crossbeam_channel::bounded(result.len());
+
+        let parallel_handler =
+            ParallelHandler::new("smartctl data", 4, move |device: (String, String)| {
+                let smart_data = get_smart_data(Path::new(&device.1), false)?;
+                tx.send((device.0, smart_data))?;
+                Ok(())
+            });
+
+        for (name, path) in device_paths.into_iter() {
+            if let Some(p) = path {
+                parallel_handler.send((name, p))?;
+            }
+        }
+
+        parallel_handler.complete()?;
+        while let Ok(msg) = rx.recv() {
+            if let Some(value) = result.get_mut(&msg.0) {
+                value.wearout = msg.1.wearout;
+                value.status = msg.1.status;
+            }
+        }
+    }
     Ok(result)
 }
 
-- 
2.39.5



_______________________________________________
pbs-devel mailing list
pbs-devel@lists.proxmox.com
https://lists.proxmox.com/cgi-bin/mailman/listinfo/pbs-devel

^ permalink raw reply	[flat|nested] 3+ messages in thread

* [pbs-devel] applied: [PATCH proxmox-backup v2 1/2] api: avoid retrieving lsblk result twice
  2024-09-17  8:05 [pbs-devel] [PATCH proxmox-backup v2 1/2] api: avoid retrieving lsblk result twice Gabriel Goller
  2024-09-17  8:05 ` [pbs-devel] [PATCH proxmox-backup v2 2/2] api: parallelize smartctl checks Gabriel Goller
@ 2024-11-12 20:18 ` Thomas Lamprecht
  1 sibling, 0 replies; 3+ messages in thread
From: Thomas Lamprecht @ 2024-11-12 20:18 UTC (permalink / raw)
  To: Proxmox Backup Server development discussion, Gabriel Goller

Am 17.09.24 um 10:05 schrieb Gabriel Goller:
> Avoid running `lsblk` twice when executing the `list_disk`
> endpoint/command. This and the various other small nits improve the
> performance of the endpoint.
> 
> Does not really fix, but is related to: #4961.
> 
> Signed-off-by: Gabriel Goller <g.goller@proxmox.com>
> ---
> 
> v2:
>  - nothing
> 
>  src/api2/node/disks/mod.rs    |  6 +++++-
>  src/tools/disks/mod.rs        | 10 +++++++---
>  src/tools/disks/smart.rs      | 10 +++-------
>  src/tools/disks/zfs.rs        | 15 ++++++++-------
>  src/tools/disks/zpool_list.rs |  2 +-
>  5 files changed, 24 insertions(+), 19 deletions(-)
> 
>

applied series, thanks!

While parallelism can also cause issues (load), this should be fine here and
it's only an implementation detail, so we always can change this back without
much hassle anyway.


_______________________________________________
pbs-devel mailing list
pbs-devel@lists.proxmox.com
https://lists.proxmox.com/cgi-bin/mailman/listinfo/pbs-devel


^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2024-11-12 20:18 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2024-09-17  8:05 [pbs-devel] [PATCH proxmox-backup v2 1/2] api: avoid retrieving lsblk result twice Gabriel Goller
2024-09-17  8:05 ` [pbs-devel] [PATCH proxmox-backup v2 2/2] api: parallelize smartctl checks Gabriel Goller
2024-11-12 20:18 ` [pbs-devel] applied: [PATCH proxmox-backup v2 1/2] api: avoid retrieving lsblk result twice Thomas Lamprecht

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
Service provided by Proxmox Server Solutions GmbH | Privacy | Legal