From 74e2bcccd5e2be04ba0387f555ce7c2fe32f34eb Mon Sep 17 00:00:00 2001
From: Matan Breizman <mbreizma@redhat.com>
Date: Wed, 2 Nov 2022 10:40:03 +0000
Subject: [PATCH] osd: Fix check_past_interval_bounds()

When getting the required past interval bounds we use
oldest_map or current pg info (lec/ec).
Before this change we set oldest_map epoch using the
osd's superblock.oldest_map.
The fix will use the max_oldest_map received with other peers
instead since a specific osd's oldest_map can lag for a while
in order to avoid large workloads.

Fixes: https://tracker.ceph.com/issues/49689

Signed-off-by: Matan Breizman <mbreizma@redhat.com>
(cherry picked from commit 0c611b362fb9cc4225f18283f74299551c2c5953)
---
 src/crimson/osd/pg.h    | 5 +++++
 src/osd/PG.cc           | 4 ++++
 src/osd/PG.h            | 1 +
 src/osd/PeeringState.cc | 4 +++-
 src/osd/PeeringState.h  | 1 +
 5 files changed, 14 insertions(+), 1 deletion(-)

diff --git a/src/crimson/osd/pg.h b/src/crimson/osd/pg.h
index 34676ee7a109..9208cc2d4764 100644
--- a/src/crimson/osd/pg.h
+++ b/src/crimson/osd/pg.h
@@ -349,6 +349,11 @@ public:
     return 0;
   }
 
+  epoch_t max_oldest_stored_osdmap() final {
+    // TODO
+    return 0;
+  }
+
   void on_backfill_reserved() final {
     recovery_handler->on_backfill_reserved();
   }
diff --git a/src/osd/PG.cc b/src/osd/PG.cc
index 5b10f146616f..fcd63807274d 100644
--- a/src/osd/PG.cc
+++ b/src/osd/PG.cc
@@ -1576,6 +1576,10 @@ epoch_t PG::oldest_stored_osdmap() {
   return osd->get_superblock().oldest_map;
 }
 
+epoch_t PG::max_oldest_stored_osdmap() {
+  return osd->get_superblock().max_oldest_map;
+}
+
 OstreamTemp PG::get_clog_info() {
   return osd->clog->info();
 }
diff --git a/src/osd/PG.h b/src/osd/PG.h
index 61adae1205eb..408e8114ef51 100644
--- a/src/osd/PG.h
+++ b/src/osd/PG.h
@@ -519,6 +519,7 @@ public:
   void clear_primary_state() override;
 
   epoch_t oldest_stored_osdmap() override;
+  epoch_t max_oldest_stored_osdmap() override;
   OstreamTemp get_clog_error() override;
   OstreamTemp get_clog_info() override;
   OstreamTemp get_clog_debug() override;
diff --git a/src/osd/PeeringState.cc b/src/osd/PeeringState.cc
index 9709f3ce1444..5fa76a7a7b23 100644
--- a/src/osd/PeeringState.cc
+++ b/src/osd/PeeringState.cc
@@ -953,7 +953,9 @@ static pair<epoch_t, epoch_t> get_required_past_interval_bounds(
 
 void PeeringState::check_past_interval_bounds() const
 {
-  auto oldest_epoch = pl->oldest_stored_osdmap();
+  // a specific OSD's oldest_map can lag for a while, therfore
+  // use the maximum MOSDMap.oldest_map received with peers.
+  auto oldest_epoch = pl->max_oldest_stored_osdmap();
   auto rpib = get_required_past_interval_bounds(
     info,
     oldest_epoch);
diff --git a/src/osd/PeeringState.h b/src/osd/PeeringState.h
index 2cc340cb9df9..d3ac117b9616 100644
--- a/src/osd/PeeringState.h
+++ b/src/osd/PeeringState.h
@@ -391,6 +391,7 @@ public:
     virtual void on_active_actmap() = 0;
     virtual void on_active_advmap(const OSDMapRef &osdmap) = 0;
     virtual epoch_t oldest_stored_osdmap() = 0;
+    virtual epoch_t max_oldest_stored_osdmap() = 0;
 
     // ============ recovery reservation notifications ==========
     virtual void on_backfill_reserved() = 0;
-- 
2.47.3