public inbox for pbs-devel@lists.proxmox.com
 help / color / mirror / Atom feed
From: Christian Ebner <c.ebner@proxmox.com>
To: "Proxmox Backup Server development discussion"
	<pbs-devel@lists.proxmox.com>,
	"Fabian Grünbichler" <f.gruenbichler@proxmox.com>
Subject: Re: [pbs-devel] [PATCH proxmox-backup v2 1/8] GC: Move S3 delete list state and logic to a dedicated struct
Date: Wed, 14 Jan 2026 09:22:12 +0100	[thread overview]
Message-ID: <3a49e618-3f42-4d05-b08a-186a9bd5a648@proxmox.com> (raw)
In-Reply-To: <1768296817.hlvzg291lq.astroid@yuna.none>

On 1/13/26 11:22 AM, Fabian Grünbichler wrote:
> On December 11, 2025 4:38 pm, Christian Ebner wrote:
>> To better keep track of the state and declutter the code on the
>> callsites, bundle the S3 delete list and it's logic by a dedicated
>> struct. Since the check for empty lists is now performed as part
>> of the deletion related methods, the callsites can get rid of that.
>>
>> Further, avoid the proxmox_time::epoch_i64() and use SystemTime and
>> Duration with their methods directly.
>>
>> Suggested-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
>> Signed-off-by: Christian Ebner <c.ebner@proxmox.com>
>> ---
>>   pbs-datastore/src/datastore.rs | 132 +++++++++++++++++++--------------
>>   1 file changed, 75 insertions(+), 57 deletions(-)
>>
>> diff --git a/pbs-datastore/src/datastore.rs b/pbs-datastore/src/datastore.rs
>> index 9c57aaac1..58fd034fc 100644
>> --- a/pbs-datastore/src/datastore.rs
>> +++ b/pbs-datastore/src/datastore.rs
>> @@ -24,7 +24,7 @@ use proxmox_sys::error::SysError;
>>   use proxmox_sys::fs::{file_read_optional_string, replace_file, CreateOptions};
>>   use proxmox_sys::linux::procfs::MountInfo;
>>   use proxmox_sys::process_locker::{ProcessLockExclusiveGuard, ProcessLockSharedGuard};
>> -use proxmox_time::{epoch_i64, TimeSpan};
> 
> we already use epoch_i64 for other calculations in this module, why not
> keep it?

Do you mean with respect to the module level import or with respect to 
using the epoch_i64() helper? As far as I see this is only used once to 
get the phase 1 start time apart from the delete list age calculations.

Not sure what we would gain from latter? As this helper wraps around the 
SystemTime anyways [0], so I opted to use that directly instead of 
converting it to the Unix epoch and doing the calculations there.

[0] 
https://rustdoc.intra.proxmox.com/trixie/packages/proxmox/src/proxmox_time/posix.rs.html#81-93

> 
>> +use proxmox_time::TimeSpan;
>>   use proxmox_worker_task::WorkerTaskContext;
>>   
>>   use pbs_api_types::{
>> @@ -64,7 +64,7 @@ const CHUNK_LOCK_TIMEOUT: Duration = Duration::from_secs(3 * 60 * 60);
>>   // s3 deletion batch size to avoid 1024 open files soft limit
>>   const S3_DELETE_BATCH_LIMIT: usize = 100;
>>   // max defer time for s3 batch deletions
>> -const S3_DELETE_DEFER_LIMIT_SECONDS: i64 = 60 * 5;
>> +const S3_DELETE_DEFER_LIMIT_SECONDS: Duration = Duration::from_secs(60 * 5);
>>   
>>   /// checks if auth_id is owner, or, if owner is a token, if
>>   /// auth_id is the user of the token
>> @@ -1689,40 +1689,8 @@ impl DataStore {
>>                   proxmox_async::runtime::block_on(s3_client.list_objects_v2(&prefix, None))
>>                       .context("failed to list chunk in s3 object store")?;
>>   
>> -            let mut delete_list = Vec::with_capacity(S3_DELETE_BATCH_LIMIT);
>> -            let mut delete_list_age = epoch_i64();
>> -
>> -            let s3_delete_batch = |delete_list: &mut Vec<(S3ObjectKey, BackupLockGuard)>,
>> -                                   s3_client: &Arc<S3Client>|
>> -             -> Result<(), Error> {
>> -                let delete_objects_result = proxmox_async::runtime::block_on(
>> -                    s3_client.delete_objects(
>> -                        &delete_list
>> -                            .iter()
>> -                            .map(|(key, _)| key.clone())
>> -                            .collect::<Vec<S3ObjectKey>>(),
>> -                    ),
>> -                )?;
>> -                if let Some(_err) = delete_objects_result.error {
>> -                    bail!("failed to delete some objects");
>> -                }
>> -                // drops all chunk flock guards
>> -                delete_list.clear();
>> -                Ok(())
>> -            };
>> -
>> -            let add_to_delete_list =
>> -                |delete_list: &mut Vec<(S3ObjectKey, BackupLockGuard)>,
>> -                 delete_list_age: &mut i64,
>> -                 key: S3ObjectKey,
>> -                 _chunk_guard: BackupLockGuard| {
>> -                    // set age based on first insertion
>> -                    if delete_list.is_empty() {
>> -                        *delete_list_age = epoch_i64();
>> -                    }
>> -                    delete_list.push((key, _chunk_guard));
>> -                };
>> -
>> +            let mut delete_list =
>> +                S3DeleteList::with_thresholds(S3_DELETE_BATCH_LIMIT, S3_DELETE_DEFER_LIMIT_SECONDS);
>>               loop {
>>                   for content in list_bucket_result.contents {
>>                       worker.check_abort()?;
>> @@ -1779,12 +1747,7 @@ impl DataStore {
>>                                               std::fs::remove_file(chunk_path)?;
>>                                           }
>>                                       }
>> -                                    add_to_delete_list(
>> -                                        &mut delete_list,
>> -                                        &mut delete_list_age,
>> -                                        content.key,
>> -                                        _chunk_guard,
>> -                                    );
>> +                                    delete_list.push(content.key, _chunk_guard);
>>                                       Ok(())
>>                                   },
>>                               )?;
>> @@ -1792,12 +1755,7 @@ impl DataStore {
>>                       } else {
>>                           gc_status.removed_chunks += 1;
>>                           gc_status.removed_bytes += content.size;
>> -                        add_to_delete_list(
>> -                            &mut delete_list,
>> -                            &mut delete_list_age,
>> -                            content.key,
>> -                            _chunk_guard,
>> -                        );
>> +                        delete_list.push(content.key, _chunk_guard);
>>                       }
>>   
>>                       chunk_count += 1;
>> @@ -1806,12 +1764,7 @@ impl DataStore {
>>                       drop(_guard);
>>   
>>                       // limit pending deletes to avoid holding too many chunk flocks
>> -                    if delete_list.len() >= S3_DELETE_BATCH_LIMIT
>> -                        || (!delete_list.is_empty()
>> -                            && epoch_i64() - delete_list_age > S3_DELETE_DEFER_LIMIT_SECONDS)
>> -                    {
>> -                        s3_delete_batch(&mut delete_list, s3_client)?;
>> -                    }
>> +                    delete_list.conditional_delete_and_drop_locks(s3_client)?;
>>                   }
>>                   // Process next batch of chunks if there is more
>>                   if list_bucket_result.is_truncated {
>> @@ -1827,9 +1780,7 @@ impl DataStore {
>>               }
>>   
>>               // delete the last batch of objects, if there are any remaining
>> -            if !delete_list.is_empty() {
>> -                s3_delete_batch(&mut delete_list, s3_client)?;
>> -            }
>> +            delete_list.delete_and_drop_locks(s3_client)?;
>>   
>>               info!("processed {chunk_count} total chunks");
>>   
>> @@ -2768,3 +2719,70 @@ impl DataStore {
>>           result
>>       }
>>   }
>> +
>> +/// Track S3 object keys to be deleted by garbage collection while holding their file lock.
>> +struct S3DeleteList {
>> +    list: Vec<(S3ObjectKey, BackupLockGuard)>,
>> +    first_entry_added: SystemTime,
>> +    age_threshold: Duration,
>> +    capacity_threshold: usize,
>> +}
>> +
>> +impl S3DeleteList {
>> +    /// Create a new list instance with given capacity and age thresholds.
>> +    fn with_thresholds(capacity_threshold: usize, age_threshold: Duration) -> Self {
>> +        Self {
>> +            first_entry_added: SystemTime::now(), // init only, updated once added
>> +            list: Vec::with_capacity(capacity_threshold),
>> +            age_threshold,
>> +            capacity_threshold,
>> +        }
>> +    }
>> +
>> +    /// Pushes the current key and backup lock guard to the list, updating the delete list age if
>> +    /// the list was empty before the insert.
>> +    fn push(&mut self, key: S3ObjectKey, guard: BackupLockGuard) {
>> +        // set age based on first insertion
>> +        if self.list.is_empty() {
>> +            self.first_entry_added = SystemTime::now();
>> +        }
>> +        self.list.push((key, guard));
>> +    }
>> +
>> +    /// Delete the objects in the list via the provided S3 client instance.
>> +    /// Clears the list contents and frees the per-chunk file locks.
>> +    fn delete_and_drop_locks(&mut self, s3_client: &Arc<S3Client>) -> Result<(), Error> {
>> +        if self.list.is_empty() {
>> +            return Ok(());
>> +        }
>> +        let delete_objects_result = proxmox_async::runtime::block_on(
>> +            s3_client.delete_objects(
>> +                &self
>> +                    .list
>> +                    .iter()
>> +                    .map(|(key, _)| key.clone())
>> +                    .collect::<Vec<S3ObjectKey>>(),
>> +            ),
>> +        )?;
>> +        if delete_objects_result.error.is_some() {
>> +            bail!("failed to delete some objects");
>> +        }
>> +        // drops all chunk flock guards
>> +        self.list.clear();
>> +        Ok(())
>> +    }
>> +
>> +    /// Delete the object stored in the list if either the list exceeds the capacity threshold or
>> +    /// the delete list age threshold.
>> +    fn conditional_delete_and_drop_locks(
>> +        &mut self,
>> +        s3_client: &Arc<S3Client>,
>> +    ) -> Result<(), Error> {
>> +        if self.list.len() >= self.capacity_threshold
>> +            || (!self.list.is_empty() && self.first_entry_added.elapsed()? > self.age_threshold)
> 
> this bails if the clock jumps backwards (further than the
> first_entry_added timestamp). maybe we should instead always delete? or
> at least add a meaningful error message/context..

Yes indeed! Deleting this in case the clock jumped to the past is 
probably the best option here, as the local chunk marker file is no 
longer present anyways at this point.

Will adapt the code accordingly.

> 
>> +        {
>> +            self.delete_and_drop_locks(s3_client)?;
>> +        }
>> +        Ok(())
>> +    }
>> +}
>> -- 
>> 2.47.3
>>
>>
>>
>> _______________________________________________
>> pbs-devel mailing list
>> pbs-devel@lists.proxmox.com
>> https://lists.proxmox.com/cgi-bin/mailman/listinfo/pbs-devel
>>
>>
>>
> 
> 
> _______________________________________________
> pbs-devel mailing list
> pbs-devel@lists.proxmox.com
> https://lists.proxmox.com/cgi-bin/mailman/listinfo/pbs-devel
> 
> 



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

  reply	other threads:[~2026-01-14  8:22 UTC|newest]

Thread overview: 23+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-12-11 15:38 [pbs-devel] [PATCH proxmox-backup v2 0/8] followups for garbage collection Christian Ebner
2025-12-11 15:38 ` [pbs-devel] [PATCH proxmox-backup v2 1/8] GC: Move S3 delete list state and logic to a dedicated struct Christian Ebner
2026-01-13 10:23   ` Fabian Grünbichler
2026-01-14  8:22     ` Christian Ebner [this message]
2026-01-14  9:18       ` Fabian Grünbichler
2025-12-11 15:38 ` [pbs-devel] [PATCH proxmox-backup v2 2/8] chunk store: rename and limit scope for chunk store iterator Christian Ebner
2025-12-11 15:38 ` [pbs-devel] [PATCH proxmox-backup v2 3/8] chunk store: invert chunk filename checks in " Christian Ebner
2026-01-13 10:23   ` Fabian Grünbichler
2026-01-14  8:37     ` Christian Ebner
2026-01-14  9:41       ` Fabian Grünbichler
2026-01-14  9:53         ` Christian Ebner
2025-12-11 15:38 ` [pbs-devel] [PATCH proxmox-backup v2 4/8] chunk store: return chunk extension and check for used marker Christian Ebner
2026-01-13 10:24   ` Fabian Grünbichler
2026-01-14  8:41     ` Christian Ebner
2025-12-11 15:38 ` [pbs-devel] [PATCH proxmox-backup v2 5/8] chunk store: refactor chunk extension parsing into dedicated helper Christian Ebner
2026-01-13 10:24   ` Fabian Grünbichler
2025-12-11 15:38 ` [pbs-devel] [PATCH proxmox-backup v2 6/8] datastore: move bad chunk touching logic to chunk store Christian Ebner
2026-01-13 10:24   ` Fabian Grünbichler
2026-01-14  8:58     ` Christian Ebner
2025-12-11 15:38 ` [pbs-devel] [PATCH proxmox-backup v2 7/8] chunk store: move next bad chunk path generator into dedicated helper Christian Ebner
2025-12-11 15:38 ` [pbs-devel] [PATCH proxmox-backup v2 8/8] chunk store: move bad chunk filename generation " Christian Ebner
2026-01-13 10:24 ` [pbs-devel] [PATCH proxmox-backup v2 0/8] followups for garbage collection Fabian Grünbichler
2026-01-14 12:33 ` [pbs-devel] superseded: " Christian Ebner

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=3a49e618-3f42-4d05-b08a-186a9bd5a648@proxmox.com \
    --to=c.ebner@proxmox.com \
    --cc=f.gruenbichler@proxmox.com \
    --cc=pbs-devel@lists.proxmox.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
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