From: Josh Durgin <jdurgin@redhat.com>
Date: Wed, 21 Jun 2017 00:29:04 +0000 (-0400)
Subject: osd/PrimaryLogPG: check whether clones missing from the cache are recovering
X-Git-Tag: v12.1.2~220^2~15
X-Git-Url: http://git-server-git.apps.pok.os.sepia.ceph.com/?a=commitdiff_plain;h=b41fa997261658003bd6bdd9ccd2665acae4cb7c;p=ceph.git

osd/PrimaryLogPG: check whether clones missing from the cache are recovering

This appears now that deletes are not processed inline from the PG log
- a clone that is missing only on a peer (due to being deleted) would
not stop rollback from promoting the clone, resulting in hitting an
assert on the replica when the promotion tried to write to the missing
object on the replica.

This only affects cache tiering due to the dependence on the
MAP_SNAP_CLONE flag in find_object_context() - missing_oid was not being checked for being
recovered, unlike the target oid for the op (in do_op()).

Signed-off-by: Josh Durgin <jdurgin@redhat.com>
---

diff --git a/src/osd/PrimaryLogPG.cc b/src/osd/PrimaryLogPG.cc
index dce2d43e6231..d8efc4f609fc 100644
--- a/src/osd/PrimaryLogPG.cc
+++ b/src/osd/PrimaryLogPG.cc
@@ -6771,8 +6771,8 @@ int PrimaryLogPG::_rollback_to(OpContext *ctx, ceph_osd_op& op)
     &rollback_to, false, false, &missing_oid);
   if (ret == -EAGAIN) {
     /* clone must be missing */
-    assert(is_missing_object(missing_oid));
-    dout(20) << "_rollback_to attempted to roll back to a missing object "
+    assert(is_degraded_or_backfilling_object(missing_oid));
+    dout(20) << "_rollback_to attempted to roll back to a missing or backfilling clone "
 	     << missing_oid << " (requested snapid: ) " << snapid << dendl;
     block_write_on_degraded_snap(missing_oid, ctx->op);
     return ret;
@@ -9890,11 +9890,16 @@ int PrimaryLogPG::find_object_context(const hobject_t& oid,
 
   ObjectContextRef obc = get_object_context(soid, false);
   if (!obc || !obc->obs.exists) {
-    dout(20) << __func__ << " missing clone " << soid << dendl;
     if (pmissing)
       *pmissing = soid;
     put_snapset_context(ssc);
-    return -ENOENT;
+    if (is_degraded_or_backfilling_object(soid)) {
+      dout(20) << __func__ << " clone is degraded or backfilling " << soid << dendl;
+      return -EAGAIN;
+    } else {
+      dout(20) << __func__ << " missing clone " << soid << dendl;
+      return -ENOENT;
+    }
   }
 
   if (!obc->ssc) {