From: Sage Weil <sage@inktank.com>
Date: Mon, 30 Dec 2013 20:57:28 +0000 (-0800)
Subject: osd/ReplicatedPG: infer snaps from head when promoting oldest clean clone
X-Git-Tag: v0.77~22^2~19
X-Git-Url: http://git-server-git.apps.pok.os.sepia.ceph.com/?a=commitdiff_plain;h=627bdead1ef40cb4cea731f4eadd416551283a01;p=ceph.git

osd/ReplicatedPG: infer snaps from head when promoting oldest clean clone

Consider:

 - base and cache have same object foo; marked clean in cache pool
 - modify + clone foo in cache pool.  foo clone is clean.
 - foo clone is evicted
 - foo clone is read, and promoted
 - we read foo@something from base pool, and get the head's content

copy-get does not provide us with a snaps list.  Instead, we use the
snap_seq from the head to infer what the snaps vector was in the cache
pool and will be in the base pool when we flush the updates to the object.

Signed-off-by: Sage Weil <sage@inktank.com>
---

diff --git a/src/osd/ReplicatedPG.cc b/src/osd/ReplicatedPG.cc
index e370e3feede2..ea862ccb7005 100644
--- a/src/osd/ReplicatedPG.cc
+++ b/src/osd/ReplicatedPG.cc
@@ -5144,7 +5144,29 @@ void ReplicatedPG::finish_promote(int r, OpRequestRef op,
     tctx->delta_stats.num_bytes += results->object_size;
     tctx->new_obs.oi.category = results->category;
     tctx->new_obs.oi.user_version = results->user_version;
-    tctx->new_obs.oi.snaps = results->snaps;
+
+    if (soid.snap != CEPH_NOSNAP) {
+      if (!results->snaps.empty()) {
+	tctx->new_obs.oi.snaps = results->snaps;
+      } else {
+	// we must have read "snap" content from the head object in
+	// the base pool.  use snap_seq to construct what snaps should
+	// be for this clone (what is was before we evicted the clean
+	// clone from this pool, and what it will be when we flush and
+	// the clone eventually happens in the base pool).
+	SnapSet& snapset = obc->ssc->snapset;
+	vector<snapid_t>::iterator p = snapset.snaps.begin();
+	while (p != snapset.snaps.end() && *p > soid.snap)
+	  ++p;
+	assert(p != snapset.snaps.end() && *p == soid.snap);
+	do {
+	  tctx->new_obs.oi.snaps.push_back(*p);
+	  ++p;
+	} while (p != snapset.snaps.end() && *p > results->snap_seq);
+      }
+      dout(20) << __func__ << " snaps " << tctx->new_obs.oi.snaps << dendl;
+      assert(!tctx->new_obs.oi.snaps.empty());
+    }
   }
 
   if (results->mirror_snapset) {