* [pbs-devel] [PATCH proxmox-backup v3] api: make prune-group a real workertask
@ 2024-03-08 13:36 Gabriel Goller
2024-03-12 10:30 ` Stefan Lendl
` (2 more replies)
0 siblings, 3 replies; 6+ messages in thread
From: Gabriel Goller @ 2024-03-08 13:36 UTC (permalink / raw)
To: pbs-devel
`prune-group` is currently not a real workertask, ie it behaves like one
but doesn't start a thread nor a task to do its work.
Changed it to start a tokio-task, so that we can delete snapshots
asynchronously. The `dry-run` feature still behaves in the same way and
returns early.
This paves the way for the new logging infra (which uses `task_local` to
define a logger) and improves performance of bigger backup-groups.
Signed-off-by: Gabriel Goller <g.goller@proxmox.com>
---
v3, thanks @Max
* remove use-task parameter, extract from serde Value
* simplify dry-run statement
* removed prune-result output
* inline variables in formatting call
v2, thanks @Fiona, @Thomas:
* use feature flag to activate, so we don't break the api
* convert the result to a structure and print it in the tasklog
* enable the feature flag in the frontend
src/api2/admin/datastore.rs | 152 ++++++++++++++++++++++--------------
www/datastore/Prune.js | 1 +
2 files changed, 96 insertions(+), 57 deletions(-)
diff --git a/src/api2/admin/datastore.rs b/src/api2/admin/datastore.rs
index a95031e7..0e55a2ca 100644
--- a/src/api2/admin/datastore.rs
+++ b/src/api2/admin/datastore.rs
@@ -944,6 +944,12 @@ pub fn verify(
type: BackupNamespace,
optional: true,
},
+ "use-task": {
+ type: bool,
+ default: false,
+ optional: true,
+ description: "Spins up an asynchronous task that does the work.",
+ },
},
},
returns: pbs_api_types::ADMIN_DATASTORE_PRUNE_RETURN_TYPE,
@@ -960,7 +966,7 @@ pub fn prune(
keep_options: KeepOptions,
store: String,
ns: Option<BackupNamespace>,
- _param: Value,
+ param: Value,
rpcenv: &mut dyn RpcEnvironment,
) -> Result<Value, Error> {
let auth_id: Authid = rpcenv.get_auth_id().unwrap().parse()?;
@@ -978,7 +984,20 @@ pub fn prune(
let worker_id = format!("{}:{}:{}", store, ns, group);
let group = datastore.backup_group(ns.clone(), group);
- let mut prune_result = Vec::new();
+ #[derive(Debug, serde::Serialize)]
+ struct PruneResult {
+ #[serde(rename = "backup-type")]
+ backup_type: BackupType,
+ #[serde(rename = "backup-id")]
+ backup_id: String,
+ #[serde(rename = "backup-time")]
+ backup_time: i64,
+ keep: bool,
+ protected: bool,
+ #[serde(skip_serializing_if = "Option::is_none")]
+ ns: Option<BackupNamespace>,
+ }
+ let mut prune_result: Vec<PruneResult> = Vec::new();
let list = group.list_backups()?;
@@ -991,78 +1010,97 @@ pub fn prune(
if dry_run {
for (info, mark) in prune_info {
let keep = keep_all || mark.keep();
-
- let mut result = json!({
- "backup-type": info.backup_dir.backup_type(),
- "backup-id": info.backup_dir.backup_id(),
- "backup-time": info.backup_dir.backup_time(),
- "keep": keep,
- "protected": mark.protected(),
- });
- let prune_ns = info.backup_dir.backup_ns();
+ let backup_dir = &info.backup_dir;
+
+ let mut result = PruneResult {
+ backup_type: backup_dir.backup_type(),
+ backup_id: backup_dir.backup_id().to_owned(),
+ backup_time: backup_dir.backup_time(),
+ keep,
+ protected: mark.protected(),
+ ns: None,
+ };
+ let prune_ns = backup_dir.backup_ns();
if !prune_ns.is_root() {
- result["ns"] = serde_json::to_value(prune_ns)?;
+ result.ns = Some(prune_ns.to_owned());
}
prune_result.push(result);
}
return Ok(json!(prune_result));
}
- // We use a WorkerTask just to have a task log, but run synchrounously
- let worker = WorkerTask::new("prune", Some(worker_id), auth_id.to_string(), true)?;
-
- if keep_all {
- task_log!(worker, "No prune selection - keeping all files.");
- } else {
- let mut opts = Vec::new();
- if !ns.is_root() {
- opts.push(format!("--ns {ns}"));
+ let prune_group = move |worker: Arc<WorkerTask>| {
+ if keep_all {
+ task_log!(worker, "No prune selection - keeping all files.");
+ } else {
+ let mut opts = Vec::new();
+ if !ns.is_root() {
+ opts.push(format!("--ns {ns}"));
+ }
+ crate::server::cli_keep_options(&mut opts, &keep_options);
+
+ task_log!(worker, "retention options: {}", opts.join(" "));
+ task_log!(
+ worker,
+ "Starting prune on {} group \"{}\"",
+ print_store_and_ns(&store, &ns),
+ group.group(),
+ );
}
- crate::server::cli_keep_options(&mut opts, &keep_options);
-
- task_log!(worker, "retention options: {}", opts.join(" "));
- task_log!(
- worker,
- "Starting prune on {} group \"{}\"",
- print_store_and_ns(&store, &ns),
- group.group(),
- );
- }
- for (info, mark) in prune_info {
- let keep = keep_all || mark.keep();
+ for (info, mark) in prune_info {
+ let keep = keep_all || mark.keep();
+ let backup_dir = &info.backup_dir;
- let backup_time = info.backup_dir.backup_time();
- let timestamp = info.backup_dir.backup_time_string();
- let group: &pbs_api_types::BackupGroup = info.backup_dir.as_ref();
+ let backup_time = backup_dir.backup_time();
+ let timestamp = backup_dir.backup_time_string();
+ let group: &pbs_api_types::BackupGroup = backup_dir.as_ref();
- let msg = format!("{}/{}/{} {}", group.ty, group.id, timestamp, mark,);
+ let msg = format!("{}/{}/{timestamp} {mark}", group.ty, group.id);
- task_log!(worker, "{}", msg);
+ task_log!(worker, "{msg}");
- prune_result.push(json!({
- "backup-type": group.ty,
- "backup-id": group.id,
- "backup-time": backup_time,
- "keep": keep,
- "protected": mark.protected(),
- }));
+ prune_result.push(PruneResult {
+ backup_type: group.ty,
+ backup_id: group.id.clone(),
+ backup_time,
+ keep,
+ protected: mark.protected(),
+ ns: None,
+ });
- if !(dry_run || keep) {
- if let Err(err) = info.backup_dir.destroy(false) {
- task_warn!(
- worker,
- "failed to remove dir {:?}: {}",
- info.backup_dir.relative_path(),
- err,
- );
+ if !keep {
+ if let Err(err) = backup_dir.destroy(false) {
+ task_warn!(
+ worker,
+ "failed to remove dir {:?}: {}",
+ backup_dir.relative_path(),
+ err,
+ );
+ }
}
}
- }
-
- worker.log_result(&Ok(()));
+ prune_result
+ };
- Ok(json!(prune_result))
+ if param["use-task"].as_bool().unwrap_or(false) {
+ let upid = WorkerTask::spawn(
+ "prune",
+ Some(worker_id),
+ auth_id.to_string(),
+ true,
+ move |worker| async move {
+ let _ = prune_group(worker.clone());
+ Ok(())
+ },
+ )?;
+ Ok(json!(upid))
+ } else {
+ let worker = WorkerTask::new("prune", Some(worker_id), auth_id.to_string(), true)?;
+ let result = prune_group(worker.clone());
+ worker.log_result(&Ok(()));
+ Ok(json!(result))
+ }
}
#[api(
diff --git a/www/datastore/Prune.js b/www/datastore/Prune.js
index 81f6927b..5752907e 100644
--- a/www/datastore/Prune.js
+++ b/www/datastore/Prune.js
@@ -52,6 +52,7 @@ Ext.define('PBS.Datastore.PruneInputPanel', {
if (me.ns && me.ns !== '') {
values.ns = me.ns;
}
+ values["use-task"] = true;
return values;
},
--
2.43.0
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [pbs-devel] [PATCH proxmox-backup v3] api: make prune-group a real workertask
2024-03-08 13:36 [pbs-devel] [PATCH proxmox-backup v3] api: make prune-group a real workertask Gabriel Goller
@ 2024-03-12 10:30 ` Stefan Lendl
2024-03-12 13:07 ` Gabriel Goller
2024-04-09 9:04 ` Gabriel Goller
2024-04-09 10:44 ` [pbs-devel] applied: " Thomas Lamprecht
2 siblings, 1 reply; 6+ messages in thread
From: Stefan Lendl @ 2024-03-12 10:30 UTC (permalink / raw)
To: pbs-devel
LGTM with a slight note below for a style preference.
> + let prune_group = move |worker: Arc<WorkerTask>| {
> + if keep_all {
> + task_log!(worker, "No prune selection - keeping all files.");
> + } else {
You don't actually need to spawn a task if you're not doing any work.
On the other hand it makes the code slightly simpler and you need the
worker, right?
> + let mut opts = Vec::new();
> + if !ns.is_root() {
> + opts.push(format!("--ns {ns}"));
> + }
> + crate::server::cli_keep_options(&mut opts, &keep_options);
> +
> + task_log!(worker, "retention options: {}", opts.join(" "));
> + task_log!(
> + worker,
> + "Starting prune on {} group \"{}\"",
> + print_store_and_ns(&store, &ns),
> + group.group(),
> + );
> }
> - crate::server::cli_keep_options(&mut opts, &keep_options);
> -
> - task_log!(worker, "retention options: {}", opts.join(" "));
> - task_log!(
> - worker,
> - "Starting prune on {} group \"{}\"",
> - print_store_and_ns(&store, &ns),
> - group.group(),
> - );
> - }
>
> - for (info, mark) in prune_info {
> - let keep = keep_all || mark.keep();
> + for (info, mark) in prune_info {
> + let keep = keep_all || mark.keep();
> + let backup_dir = &info.backup_dir;
You wouldn't have to handle keep_all here.
You're handling dry_run separatly already as well.
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [pbs-devel] [PATCH proxmox-backup v3] api: make prune-group a real workertask
2024-03-12 10:30 ` Stefan Lendl
@ 2024-03-12 13:07 ` Gabriel Goller
2024-04-02 11:14 ` Stefan Lendl
0 siblings, 1 reply; 6+ messages in thread
From: Gabriel Goller @ 2024-03-12 13:07 UTC (permalink / raw)
To: Proxmox Backup Server development discussion
On Tue Mar 12, 2024 at 11:30 AM CET, Stefan Lendl wrote:
>
> LGTM with a slight note below for a style preference.
>
> > + let prune_group = move |worker: Arc<WorkerTask>| {
> > + if keep_all {
> > + task_log!(worker, "No prune selection - keeping all files.");
> > + } else {
>
> You don't actually need to spawn a task if you're not doing any work.
> On the other hand it makes the code slightly simpler and you need the
> worker, right?
IMO starting a worker is better, because the user sees that something is
happening... I mean we could also immediately return and show a popup in
the frontend, but that won't work when the user isn't logged
in/currently looking at the page. This way he simply sees a prune job in
the task log which prints 'No prune selection - keeping all files'.
> > [snip]
> > - for (info, mark) in prune_info {
> > - let keep = keep_all || mark.keep();
>
> > + for (info, mark) in prune_info {
> > + let keep = keep_all || mark.keep();
> > + let backup_dir = &info.backup_dir;
>
> You wouldn't have to handle keep_all here.
>
> You're handling dry_run separatly already as well.
Hmm I think I need it here though, note that this is not inside the
if/else statement above. Even if we have keep_all we still want to go
through all the snapshots and mark them (+ output mark) I guess.
Thanks for the review!
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [pbs-devel] [PATCH proxmox-backup v3] api: make prune-group a real workertask
2024-03-12 13:07 ` Gabriel Goller
@ 2024-04-02 11:14 ` Stefan Lendl
0 siblings, 0 replies; 6+ messages in thread
From: Stefan Lendl @ 2024-04-02 11:14 UTC (permalink / raw)
To: Gabriel Goller, Proxmox Backup Server development discussion
"Gabriel Goller" <g.goller@proxmox.com> writes:
> On Tue Mar 12, 2024 at 11:30 AM CET, Stefan Lendl wrote:
>>
>> LGTM with a slight note below for a style preference.
>>
>> > + let prune_group = move |worker: Arc<WorkerTask>| {
>> > + if keep_all {
>> > + task_log!(worker, "No prune selection - keeping all files.");
>> > + } else {
>>
>> You don't actually need to spawn a task if you're not doing any work.
>> On the other hand it makes the code slightly simpler and you need the
>> worker, right?
>
> IMO starting a worker is better, because the user sees that something is
> happening... I mean we could also immediately return and show a popup in
> the frontend, but that won't work when the user isn't logged
> in/currently looking at the page. This way he simply sees a prune job in
> the task log which prints 'No prune selection - keeping all files'.
>
Makes sense!
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [pbs-devel] [PATCH proxmox-backup v3] api: make prune-group a real workertask
2024-03-08 13:36 [pbs-devel] [PATCH proxmox-backup v3] api: make prune-group a real workertask Gabriel Goller
2024-03-12 10:30 ` Stefan Lendl
@ 2024-04-09 9:04 ` Gabriel Goller
2024-04-09 10:44 ` [pbs-devel] applied: " Thomas Lamprecht
2 siblings, 0 replies; 6+ messages in thread
From: Gabriel Goller @ 2024-04-09 9:04 UTC (permalink / raw)
To: Proxmox Backup Server development discussion
bump, this should still apply on master.
^ permalink raw reply [flat|nested] 6+ messages in thread
* [pbs-devel] applied: [PATCH proxmox-backup v3] api: make prune-group a real workertask
2024-03-08 13:36 [pbs-devel] [PATCH proxmox-backup v3] api: make prune-group a real workertask Gabriel Goller
2024-03-12 10:30 ` Stefan Lendl
2024-04-09 9:04 ` Gabriel Goller
@ 2024-04-09 10:44 ` Thomas Lamprecht
2 siblings, 0 replies; 6+ messages in thread
From: Thomas Lamprecht @ 2024-04-09 10:44 UTC (permalink / raw)
To: Proxmox Backup Server development discussion, Gabriel Goller
Am 08/03/2024 um 14:36 schrieb Gabriel Goller:
> `prune-group` is currently not a real workertask, ie it behaves like one
> but doesn't start a thread nor a task to do its work.
>
> Changed it to start a tokio-task, so that we can delete snapshots
> asynchronously. The `dry-run` feature still behaves in the same way and
> returns early.
>
> This paves the way for the new logging infra (which uses `task_local` to
> define a logger) and improves performance of bigger backup-groups.
>
> Signed-off-by: Gabriel Goller <g.goller@proxmox.com>
> ---
>
> v3, thanks @Max
> * remove use-task parameter, extract from serde Value
> * simplify dry-run statement
> * removed prune-result output
> * inline variables in formatting call
>
> v2, thanks @Fiona, @Thomas:
> * use feature flag to activate, so we don't break the api
> * convert the result to a structure and print it in the tasklog
> * enable the feature flag in the frontend
>
>
> src/api2/admin/datastore.rs | 152 ++++++++++++++++++++++--------------
> www/datastore/Prune.js | 1 +
> 2 files changed, 96 insertions(+), 57 deletions(-)
>
>
applied, thanks!
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2024-04-09 10:44 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2024-03-08 13:36 [pbs-devel] [PATCH proxmox-backup v3] api: make prune-group a real workertask Gabriel Goller
2024-03-12 10:30 ` Stefan Lendl
2024-03-12 13:07 ` Gabriel Goller
2024-04-02 11:14 ` Stefan Lendl
2024-04-09 9:04 ` Gabriel Goller
2024-04-09 10:44 ` [pbs-devel] applied: " Thomas Lamprecht
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox