From 24fd8d6573c8e74a80cbddfa722b749595331da1 Mon Sep 17 00:00:00 2001
From: Sage Weil <sage@redhat.com>
Date: Fri, 27 Jul 2018 08:58:24 -0500
Subject: [PATCH] osd/PG: implement merge_from

This is the building block that smooshes multiple PGs back into one.  The
resulting combination PG will have no PG log.  That means the sources
need to be clean and quiesced or else the result will end up being
marked incomplete.

Signed-off-by: Sage Weil <sage@redhat.com>
---
 src/osd/PG.cc | 93 +++++++++++++++++++++++++++++++++++++++++++++++++++
 src/osd/PG.h  |  2 ++
 2 files changed, 95 insertions(+)

diff --git a/src/osd/PG.cc b/src/osd/PG.cc
index fa4a197d7dfa9..ae881ad6307d0 100644
--- a/src/osd/PG.cc
+++ b/src/osd/PG.cc
@@ -2640,6 +2640,99 @@ void PG::finish_split_stats(const object_stat_sum_t& stats, ObjectStore::Transac
   write_if_dirty(*t);
 }
 
+void PG::merge_from(map<spg_t,PGRef>& sources, RecoveryCtx *rctx,
+		    unsigned split_bits)
+{
+  dout(10) << __func__ << " from " << sources << " split_bits " << split_bits
+	   << dendl;
+  bool incomplete = false;
+  if (info.last_complete != info.last_update ||
+      info.is_incomplete() ||
+      info.dne()) {
+    dout(10) << __func__ << " target incomplete" << dendl;
+    incomplete = true;
+  }
+
+  PGLogEntryHandler handler{this, rctx->transaction};
+  pg_log.roll_forward(&handler);
+
+  info.last_complete = info.last_update;  // to fake out trim()
+  pg_log.reset_recovery_pointers();
+  pg_log.trim(info.last_update, info);
+
+  vector<PGLog*> log_from;
+  for (auto& i : sources) {
+    auto& source = i.second;
+    if (!source) {
+      dout(10) << __func__ << " source " << i.first << " missing" << dendl;
+      incomplete = true;
+      continue;
+    }
+    if (source->info.last_complete != source->info.last_update ||
+	source->info.is_incomplete() ||
+	source->info.dne()) {
+      dout(10) << __func__ << " source " << source->pg_id << " incomplete"
+	       << dendl;
+      incomplete = true;
+    }
+
+    // prepare log
+    PGLogEntryHandler handler{source.get(), rctx->transaction};
+    source->pg_log.roll_forward(&handler);
+    source->info.last_complete = source->info.last_update;  // to fake out trim()
+    source->pg_log.reset_recovery_pointers();
+    source->pg_log.trim(source->info.last_update, source->info);
+    log_from.push_back(&source->pg_log);
+
+    // wipe out source's pgmeta
+    rctx->transaction->remove(source->coll, source->pgmeta_oid);
+
+    // merge (and destroy source collection)
+    rctx->transaction->merge_collection(source->coll, coll, split_bits);
+
+    // combine stats
+    info.stats.add(source->info.stats);
+
+    // pull up last_update
+    info.last_update = std::max(info.last_update, source->info.last_update);
+
+    // adopt source's PastIntervals if target has none.  we can do this since
+    // pgp_num has been reduced prior to the merge, so the OSD mappings for
+    // the PGs are identical.
+    if (past_intervals.empty() && !source->past_intervals.empty()) {
+      dout(10) << __func__ << " taking source's past_intervals" << dendl;
+      past_intervals = source->past_intervals;
+    }
+  }
+
+  // merge_collection does this, but maybe all of our sources were missing.
+  rctx->transaction->collection_set_bits(coll, split_bits);
+
+  info.last_complete = info.last_update;
+  info.log_tail = info.last_update;
+  if (incomplete) {
+    info.last_backfill = hobject_t();
+  }
+
+  snap_mapper.update_bits(split_bits);
+
+  // merge logs
+  pg_log.merge_from(log_from, info.last_update);
+
+  // make sure we have a meaningful last_epoch_started/clean (if we were a
+  // placeholder)
+  if (info.last_epoch_started == 0) {
+    info.history.last_epoch_started =
+      info.history.last_epoch_clean = past_intervals.get_bounds().first;
+    dout(10) << __func__
+	     << " set last_epoch_started/clean based on past intervals"
+	     << dendl;
+  }
+
+  dirty_info = true;
+  dirty_big_info = true;
+}
+
 void PG::add_backoff(SessionRef s, const hobject_t& begin, const hobject_t& end)
 {
   ConnectionRef con = s->con;
diff --git a/src/osd/PG.h b/src/osd/PG.h
index 162141c71c52d..259cf6fdca072 100644
--- a/src/osd/PG.h
+++ b/src/osd/PG.h
@@ -398,6 +398,8 @@ public:
     const pg_pool_t *pool,
     ObjectStore::Transaction *t) = 0;
   void split_into(pg_t child_pgid, PG *child, unsigned split_bits);
+  void merge_from(map<spg_t,PGRef>& sources, RecoveryCtx *rctx,
+		  unsigned split_bits);
   void finish_split_stats(const object_stat_sum_t& stats, ObjectStore::Transaction *t);
 
   void scrub(epoch_t queued, ThreadPool::TPHandle &handle);
-- 
2.39.5