From 627bdead1ef40cb4cea731f4eadd416551283a01 Mon Sep 17 00:00:00 2001 From: Sage Weil Date: Mon, 30 Dec 2013 12:57:28 -0800 Subject: [PATCH] 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 --- src/osd/ReplicatedPG.cc | 24 +++++++++++++++++++++++- 1 file changed, 23 insertions(+), 1 deletion(-) diff --git a/src/osd/ReplicatedPG.cc b/src/osd/ReplicatedPG.cc index e370e3feede..ea862ccb700 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::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) {