[pbs-devel] [PATCH proxmox-backup 7/9] datastore: conditionally use custom GC wait period if set

[Christian Ebner]

Use the user configured wait period over the default 24h 5m wait
period if explicitly set, but only if the atime check was enabled
and succeeded. If the atime update check was not enabled, fallback
to use the default, as otherwise there is potential for data loss
during phase 2 of garbage collection.

Signed-off-by: Christian Ebner <c.ebner at proxmox.com>
---
changes since version 1:
- not present in previous version

 pbs-datastore/src/chunk_store.rs | 10 +++++++++-
 pbs-datastore/src/datastore.rs   | 15 ++++++++++++++-
 2 files changed, 23 insertions(+), 2 deletions(-)

diff --git a/pbs-datastore/src/chunk_store.rs b/pbs-datastore/src/chunk_store.rs
index be57e3c87..893d243aa 100644
--- a/pbs-datastore/src/chunk_store.rs
+++ b/pbs-datastore/src/chunk_store.rs
@@ -369,6 +369,7 @@ impl ChunkStore {
         &self,
         oldest_writer: i64,
         phase1_start_time: i64,
+        wait_period: Option<usize>,
         status: &mut GarbageCollectionStatus,
         worker: &dyn WorkerTaskContext,
     ) -> Result<(), Error> {
@@ -378,7 +379,14 @@ impl ChunkStore {
         use nix::sys::stat::fstatat;
         use nix::unistd::{unlinkat, UnlinkatFlags};
 
-        let mut min_atime = phase1_start_time - 3600 * 24; // at least 24h (see mount option relatime)
+        let mut min_atime = if let Some(wait_period) = wait_period {
+            // account for the 5 min safety gap subtracted below,
+            // wait period is limited to a minimum of 5 min by api schema
+            phase1_start_time + 300 - wait_period as i64 * 60
+        } else {
+            // at least 24h (see mount option relatime)
+            phase1_start_time - 3600 * 24
+        };
 
         if oldest_writer < min_atime {
             min_atime = oldest_writer;
diff --git a/pbs-datastore/src/datastore.rs b/pbs-datastore/src/datastore.rs
index 9b01790cc..ee38e10bb 100644
--- a/pbs-datastore/src/datastore.rs
+++ b/pbs-datastore/src/datastore.rs
@@ -1174,12 +1174,24 @@ impl DataStore {
                 DatastoreTuning::API_SCHEMA
                     .parse_property_string(gc_store_config.tuning.as_deref().unwrap_or(""))?,
             )?;
-            if tuning.gc_atime_check.unwrap_or(true) {
+            let gc_atime_check = tuning.gc_atime_check.unwrap_or(true);
+            if gc_atime_check {
                 self.inner.chunk_store.check_fs_atime_update()?;
                 info!("Filesystem atime update check successful.");
             } else {
                 info!("Filesystem atime update check disabled by datastore tuning options.");
             }
+            let mut wait_period = None;
+            if let Some(period) = tuning.gc_wait_period {
+                if gc_atime_check {
+                    info!("Using GC wait period of {period}m.");
+                    wait_period = Some(period);
+                } else {
+                    warn!("Ignoring GC wait period of {period}m, because atime check is disabled.");
+                }
+            } else {
+                info!("Using default GC wait period of 24h 5m");
+            }
 
             info!("Start GC phase1 (mark used chunks)");
 
@@ -1189,6 +1201,7 @@ impl DataStore {
             self.inner.chunk_store.sweep_unused_chunks(
                 oldest_writer,
                 phase1_start_time,
+                wait_period,
                 &mut gc_status,
                 worker,
             )?;
-- 
2.39.5