From: Samuel Just <sam.just@inktank.com>
Date: Tue, 26 Aug 2014 21:13:57 +0000 (-0700)
Subject: ReplicatedPG:start_flush send a second delete
X-Git-Tag: v0.85~3^2
X-Git-Url: http://git-server-git.apps.pok.os.sepia.ceph.com/?a=commitdiff_plain;h=refs%2Fpull%2F2348%2Fhead;p=ceph.git

ReplicatedPG:start_flush send a second delete

Suppose we start with the following in the cache pool:

30:[29,21,20,15,10,4]:[22(21), 15(15,10), 4(4)]+head

The object doesn't exist at 29 or 20.

First, we flush 4 leaving the backing pool with:

3:[]+head

Then, we begin to flush 15 with a delete with snapc 4:[4] leaving the
backing pool with:

4:[4]:[4(4)]

Then, we finish flushing 15 with snapc 9:[4] with leaving the backing
pool with:

9:[4]:[4(4)]+head

Next, snaps 10 and 15 are removed causing clone 10 to be removed leaving
the cache with:

30:[29,21,20,4]:[22(21),4(4)]+head

We next begin to flush 22 by sending a delete with snapc 4(4) since
prev_snapc is 4 <---------- here is the bug

The backing pool ignores this request since 4 < 9 (ORDERSNAP) leaving it
with:

9:[4]:[4(4)]

Then, we complete flushing 22 with snapc 19:[4] leaving the backing pool
with:

19:[4]:[4(4)]+head

Then, we begin to flush head by deleting with snapc 22:[21,20,4] leaving
the backing pool with:

22[21,20,4]:[22(21,20), 4(4)]

Finally, we flush head leaving the backing pool with:

30:[29,21,20,4]:[22(21*,20*),4(4)]+head

When we go to flush clone 22, all we know is that 22 is dirty, has snaps
[21], and 4 is clean. As part of flushing 22, we need to do two things:
1) Ensure that the current head is cloned as cloneid 4 with snaps [4] by
sending a delete at snapc 4:[4].
2) Flush the data at snap sequence < 21 by sending a copyfrom with snapc
20:[20,4].

Unfortunately, it is possible that 1, 1&2, or 1 and part of the flush
process for some other now non-existent clone have already been
performed.  Because of that, between 1) and 2), we need to send
a second delete ensuring that the object does not exist at 20.

Fixes: #9054
Backport: firefly
Signed-off-by: Samuel Just <sam.just@inktank.com>
---

diff --git a/src/osd/ReplicatedPG.cc b/src/osd/ReplicatedPG.cc
index b18ddeb8ed5b..eb4ea5606ebb 100644
--- a/src/osd/ReplicatedPG.cc
+++ b/src/osd/ReplicatedPG.cc
@@ -6263,7 +6263,19 @@ int ReplicatedPG::start_flush(
     cancel_flush(fop, false);
   }
 
-  // construct a SnapContext appropriate for this clone/head
+  /**
+   * In general, we need to send two deletes and a copyfrom.
+   * Consider snapc 10:[10, 9, 8, 4, 3, 2]:[10(10, 9), 4(4,3,2)]
+   * where 4 is marked as clean.  To flush 10, we have to:
+   * 1) delete 4:[4,3,2] -- ensure head is created at cloneid 4
+   * 2) delete (8-1):[4,3,2] -- ensure that the object does not exist at 8
+   * 3) copyfrom 8:[8,4,3,2] -- flush object excluding snap 8
+   *
+   * The second delete is required in case at some point in the past
+   * there had been a clone 7(7,6), which we had flushed.  Without
+   * the second delete, the object would appear in the base pool to
+   * have existed.
+   */
   SnapContext dsnapc;
   dsnapc.seq = 0;
   SnapContext snapc;
@@ -6295,17 +6307,15 @@ int ReplicatedPG::start_flush(
     vector<snapid_t>::iterator dnewest = p;
 
     // we may need to send a delete first
-    while (p != snapset.snaps.end() && *p > prev_snapc)
-      ++p;
-    dsnapc.snaps = vector<snapid_t>(p, snapset.snaps.end());
+    if (prev_snapc + 1 < *dnewest) {
+      while (p != snapset.snaps.end() && *p > prev_snapc)
+	++p;
+      dsnapc.snaps = vector<snapid_t>(p, snapset.snaps.end());
 
-    if (p == dnewest) {
-      // no snaps between the oldest in this clone and prev_snapc
-      snapc.seq = prev_snapc;
-    } else {
-      // snaps between oldest in this clone and prev_snapc, send delete
       dsnapc.seq = prev_snapc;
       snapc.seq = oi.snaps.back() - 1;
+    } else {
+      snapc.seq = prev_snapc;
     }
   }
 
@@ -6328,6 +6338,40 @@ int ReplicatedPG::start_flush(
       NULL,
       NULL /* no callback, we'll rely on the ordering w.r.t the next op */);
     osd->objecter_lock.Unlock();
+
+    // do we need to send the second delete?
+    SnapContext dsnapc2;
+    vector<snapid_t>::reverse_iterator rp = snapset.snaps.rbegin();
+
+    // advance rp to the smallest snap not contained by the last flushed clone
+    while (rp != snapset.snaps.rend() && *rp <= dsnapc.seq)
+      ++rp;
+
+    // set dnsnapc2.seq to be the snap prior to that snap (the object did not
+    // exist at *rq, so it must have been deleted prior to that).
+    dsnapc2.seq = (rp == snapset.snaps.rend()) ? snapset.seq : *rp;
+    if (dsnapc2.seq > 0)
+      dsnapc2.seq.val -= 1;
+
+    if (dsnapc2.seq != dsnapc.seq) {
+      dsnapc2.snaps = dsnapc.snaps;
+
+      ObjectOperation o2;
+      o2.remove();
+      osd->objecter_lock.Lock();
+      osd->objecter->mutate(
+	soid.oid,
+	base_oloc,
+	o2,
+	dsnapc2,
+	oi.mtime,
+	(CEPH_OSD_FLAG_IGNORE_OVERLAY |
+	 CEPH_OSD_FLAG_ORDERSNAP |
+	 CEPH_OSD_FLAG_ENFORCE_SNAPC),
+	NULL,
+	NULL /* no callback, we'll rely on the ordering w.r.t the next op */);
+      osd->objecter_lock.Unlock();
+    }
   }
 
   FlushOpRef fop(new FlushOp);