clusters/afr: Move root entry heal flow to SHD

Summary:
- Improves upon D2387001 by moving the "forced" root gfid heal to the SHDs
- Removed code which forced NFSd/FUSE clients through the entry heal for
  the root GFID, this will make them spin up just as fast as prior to D2387001 (i.e. instantly)

Porting note: mostly inapplicable in 3.8, only one non-test change survived

Test Plan: - Must pass tests/bugs/fb8149516.t

Reviewers: dph, moox, sshreyas

Reviewed By: sshreyas

Differential Revision: https://phabricator.fb.com/D2722239

Change-Id: I35f5827df6ead1bb0ff886ca0adabb2add2e7163
Signed-off-by: Jeff Darcy <jdarcy@fb.com>
Reviewed-on: https://review.gluster.org/18259
Reviewed-by: Jeff Darcy <jeff@pl.atyp.us>
Tested-by: Jeff Darcy <jeff@pl.atyp.us>
CentOS-regression: Gluster Build System <jenkins@build.gluster.org>
Smoke: Gluster Build System <jenkins@build.gluster.org>

2 files changed, 10 insertions, 6 deletions

diff --git a/tests/bugs/fb8149516.t b/tests/bugs/fb8149516.t
index 3679ab8f93a..e7f87ba29aa 100644
--- a/tests/bugs/fb8149516.t
+++ b/tests/bugs/fb8149516.t
@@ -10,40 +10,44 @@ TEST pidof glusterd
 TEST $CLI volume create $V0 replica 3 $H0:$B0/${V0}{0,1,2}
 TEST $CLI volume set $V0 cluster.read-subvolume-index 2
 TEST $CLI volume set $V0 cluster.background-self-heal-count 0
-TEST $CLI volume set $V0 cluster.heal-timeout 30
+TEST $CLI volume set $V0 cluster.heal-timeout 20
 TEST $CLI volume set $V0 cluster.choose-local off
 TEST $CLI volume set $V0 cluster.entry-self-heal off
 TEST $CLI volume set $V0 cluster.data-self-heal off
 TEST $CLI volume set $V0 cluster.metadata-self-heal off
+TEST $CLI volume set $V0 cluster.shd-max-threads 1
 TEST $CLI volume set $V0 nfs.disable off
 TEST $CLI volume set $V0 nfs.write-size 524288
 TEST $CLI volume set $V0 nfs.read-size 524288
 TEST $CLI volume start $V0
 TEST glusterfs --volfile-id=/$V0 --volfile-server=$H0 $M0 --attribute-timeout=0 --entry-timeout=0
+mkdir $M0/dir
 cd $M0
 for i in {1..10}
 do
         dd if=/dev/urandom of=testfile$i bs=1M count=1 2>/dev/null
+        dd if=/dev/urandom of=dir/testfile$i bs=1M count=1 2>/dev/null
 done
 cd ~
 TEST kill_brick $V0 $H0 $B0/${V0}2
 TEST rm -rf $B0/${V0}2/testfile*
+TEST rm -rf $B0/${V0}2/dir
 TEST rm -rf $B0/${V0}2/.glusterfs
 
 TEST $CLI volume stop $V0
 sleep 5
 TEST $CLI volume start $V0
 EXPECT_WITHIN 20 "1" afr_child_up_status_in_shd $V0 2
-EXPECT_WITHIN 30 "0" get_pending_heal_count $V0
+EXPECT_WITHIN 90 "0" get_pending_heal_count $V0
 
 # Verify we see all ten files when ls'ing, without the patch this should
 # return no files and fail.
 how_many_files () {
-        \ls $M0 2> /dev/null | wc -l
+        find $M0 -type f 2> /dev/null | wc -l
 }
 # Once it's triggered, entry self-heal happens asynchronously and might take
 # a long-ish time before the missing entries appear, but it does complete.
 # To accommodate that, retry for a little while.
-EXPECT_WITHIN 20 "10" how_many_files
+EXPECT_WITHIN 20 "20" how_many_files
 
 cleanup
diff --git a/xlators/cluster/afr/src/afr-common.c b/xlators/cluster/afr/src/afr-common.c
index c7657dd5a9f..7425b6688f2 100644
--- a/xlators/cluster/afr/src/afr-common.c
+++ b/xlators/cluster/afr/src/afr-common.c
@@ -1086,7 +1086,7 @@ int
 afr_inode_refresh_done (call_frame_t *frame, xlator_t *this, int error)
 {
 	afr_private_t *priv = NULL;
-    call_frame_t *heal_frame = NULL;
+        call_frame_t *heal_frame = NULL;
 	afr_local_t *local = NULL;
         gf_boolean_t start_heal = _gf_false;
         afr_local_t *heal_local = NULL;
@@ -1108,7 +1108,7 @@ afr_inode_refresh_done (call_frame_t *frame, xlator_t *this, int error)
 	err = afr_inode_refresh_err (frame, this);
 
         if ((ret && afr_selfheal_enabled (this)) ||
-            (!priv->shd.iamshd && (priv->did_discovery == _gf_false) &&
+            (priv->shd.iamshd &&
              AFR_IS_ROOT_GFID (local->refreshinode->gfid))) {
                 heal_frame = copy_frame (frame);
                 if (!heal_frame)