|
|
233933 |
From 6565749c95e90f360a994bde1416cffd22cd8ce9 Mon Sep 17 00:00:00 2001
|
|
|
233933 |
From: N Balachandran <nbalacha@redhat.com>
|
|
|
233933 |
Date: Mon, 25 Mar 2019 15:56:56 +0530
|
|
|
233933 |
Subject: [PATCH 126/141] cluster/dht: refactor dht lookup functions
|
|
|
233933 |
|
|
|
233933 |
Part 1: refactor the dht_lookup_dir_cbk
|
|
|
233933 |
and dht_selfheal_directory functions.
|
|
|
233933 |
Added a simple dht selfheal directory test
|
|
|
233933 |
|
|
|
233933 |
upstream: https://review.gluster.org/#/c/glusterfs/+/22407/
|
|
|
233933 |
> Change-Id: I1410c26359e3c14b396adbe751937a52bd2fcff9
|
|
|
233933 |
> updates: bz#1590385
|
|
|
233933 |
|
|
|
233933 |
Change-Id: Idd0a7df7122d634c371ecf30c0dbb94dc6063416
|
|
|
233933 |
BUG: 1703897
|
|
|
233933 |
Signed-off-by: N Balachandran <nbalacha@redhat.com>
|
|
|
233933 |
Reviewed-on: https://code.engineering.redhat.com/gerrit/169037
|
|
|
233933 |
Tested-by: RHGS Build Bot <nigelb@redhat.com>
|
|
|
233933 |
Reviewed-by: Susant Palai <spalai@redhat.com>
|
|
|
233933 |
Reviewed-by: Atin Mukherjee <amukherj@redhat.com>
|
|
|
233933 |
---
|
|
|
233933 |
tests/basic/distribute/dir-heal.t | 145 +++++++++++++++++++++++++++
|
|
|
233933 |
xlators/cluster/dht/src/dht-common.c | 178 +++++++++++++++------------------
|
|
|
233933 |
xlators/cluster/dht/src/dht-selfheal.c | 65 +++++++-----
|
|
|
233933 |
3 files changed, 264 insertions(+), 124 deletions(-)
|
|
|
233933 |
create mode 100644 tests/basic/distribute/dir-heal.t
|
|
|
233933 |
|
|
|
233933 |
diff --git a/tests/basic/distribute/dir-heal.t b/tests/basic/distribute/dir-heal.t
|
|
|
233933 |
new file mode 100644
|
|
|
233933 |
index 0000000..851f765
|
|
|
233933 |
--- /dev/null
|
|
|
233933 |
+++ b/tests/basic/distribute/dir-heal.t
|
|
|
233933 |
@@ -0,0 +1,145 @@
|
|
|
233933 |
+#!/bin/bash
|
|
|
233933 |
+
|
|
|
233933 |
+. $(dirname $0)/../../include.rc
|
|
|
233933 |
+. $(dirname $0)/../../volume.rc
|
|
|
233933 |
+. $(dirname $0)/../../nfs.rc
|
|
|
233933 |
+. $(dirname $0)/../../common-utils.rc
|
|
|
233933 |
+
|
|
|
233933 |
+# Test 1 overview:
|
|
|
233933 |
+# ----------------
|
|
|
233933 |
+#
|
|
|
233933 |
+# 1. Kill one brick of the volume.
|
|
|
233933 |
+# 2. Create directories and change directory properties.
|
|
|
233933 |
+# 3. Bring up the brick and access the directory
|
|
|
233933 |
+# 4. Check the permissions and xattrs on the backend
|
|
|
233933 |
+
|
|
|
233933 |
+cleanup
|
|
|
233933 |
+
|
|
|
233933 |
+TEST glusterd
|
|
|
233933 |
+TEST pidof glusterd
|
|
|
233933 |
+
|
|
|
233933 |
+TEST $CLI volume create $V0 $H0:$B0/$V0-{1..3}
|
|
|
233933 |
+TEST $CLI volume start $V0
|
|
|
233933 |
+
|
|
|
233933 |
+# We want the lookup to reach DHT
|
|
|
233933 |
+TEST $CLI volume set $V0 performance.stat-prefetch off
|
|
|
233933 |
+
|
|
|
233933 |
+# Mount using FUSE , kill a brick and create directories
|
|
|
233933 |
+TEST glusterfs --entry-timeout=0 --attribute-timeout=0 -s $H0 --volfile-id $V0 $M0
|
|
|
233933 |
+
|
|
|
233933 |
+ls $M0/
|
|
|
233933 |
+cd $M0
|
|
|
233933 |
+
|
|
|
233933 |
+TEST kill_brick $V0 $H0 $B0/$V0-1
|
|
|
233933 |
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "0" brick_up_status $V0 $H0 $B0/$V0-1
|
|
|
233933 |
+
|
|
|
233933 |
+TEST mkdir dir{1..4}
|
|
|
233933 |
+
|
|
|
233933 |
+# No change for dir1
|
|
|
233933 |
+# Change permissions for dir2
|
|
|
233933 |
+# Set xattr on dir3
|
|
|
233933 |
+# Change permissions and set xattr on dir4
|
|
|
233933 |
+
|
|
|
233933 |
+TEST chmod 777 $M0/dir2
|
|
|
233933 |
+
|
|
|
233933 |
+TEST setfattr -n "user.test" -v "test" $M0/dir3
|
|
|
233933 |
+
|
|
|
233933 |
+TEST chmod 777 $M0/dir4
|
|
|
233933 |
+TEST setfattr -n "user.test" -v "test" $M0/dir4
|
|
|
233933 |
+
|
|
|
233933 |
+
|
|
|
233933 |
+# Start all bricks
|
|
|
233933 |
+
|
|
|
233933 |
+TEST $CLI volume start $V0 force
|
|
|
233933 |
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "1" brick_up_status $V0 $H0 $B0/$V0-1
|
|
|
233933 |
+
|
|
|
233933 |
+#$CLI volume status
|
|
|
233933 |
+
|
|
|
233933 |
+# It takes a while for the client to reconnect to the brick
|
|
|
233933 |
+sleep 5
|
|
|
233933 |
+
|
|
|
233933 |
+stat $M0/dir* > /dev/null
|
|
|
233933 |
+
|
|
|
233933 |
+# Check that directories have been created on the brick that was killed
|
|
|
233933 |
+
|
|
|
233933 |
+TEST ls $B0/$V0-1/dir1
|
|
|
233933 |
+
|
|
|
233933 |
+TEST ls $B0/$V0-1/dir2
|
|
|
233933 |
+EXPECT "777" stat -c "%a" $B0/$V0-1/dir2
|
|
|
233933 |
+
|
|
|
233933 |
+TEST ls $B0/$V0-1/dir3
|
|
|
233933 |
+EXPECT "test" getfattr -n "user.test" --absolute-names --only-values $B0/$V0-1/dir3
|
|
|
233933 |
+
|
|
|
233933 |
+
|
|
|
233933 |
+TEST ls $B0/$V0-1/dir4
|
|
|
233933 |
+EXPECT "777" stat -c "%a" $B0/$V0-1/dir4
|
|
|
233933 |
+EXPECT "test" getfattr -n "user.test" --absolute-names --only-values $B0/$V0-1/dir4
|
|
|
233933 |
+
|
|
|
233933 |
+
|
|
|
233933 |
+TEST rm -rf $M0/*
|
|
|
233933 |
+
|
|
|
233933 |
+cd
|
|
|
233933 |
+
|
|
|
233933 |
+EXPECT_WITHIN $UMOUNT_TIMEOUT "Y" force_umount $M0
|
|
|
233933 |
+
|
|
|
233933 |
+
|
|
|
233933 |
+# Test 2 overview:
|
|
|
233933 |
+# ----------------
|
|
|
233933 |
+# 1. Create directories with all bricks up.
|
|
|
233933 |
+# 2. Kill a brick and change directory properties and set user xattr.
|
|
|
233933 |
+# 2. Bring up the brick and access the directory
|
|
|
233933 |
+# 3. Check the permissions and xattrs on the backend
|
|
|
233933 |
+
|
|
|
233933 |
+
|
|
|
233933 |
+TEST glusterfs --entry-timeout=0 --attribute-timeout=0 -s $H0 --volfile-id $V0 $M0
|
|
|
233933 |
+
|
|
|
233933 |
+ls $M0/
|
|
|
233933 |
+cd $M0
|
|
|
233933 |
+TEST mkdir dir{1..4}
|
|
|
233933 |
+
|
|
|
233933 |
+TEST kill_brick $V0 $H0 $B0/$V0-1
|
|
|
233933 |
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "0" brick_up_status $V0 $H0 $B0/$V0-1
|
|
|
233933 |
+
|
|
|
233933 |
+# No change for dir1
|
|
|
233933 |
+# Change permissions for dir2
|
|
|
233933 |
+# Set xattr on dir3
|
|
|
233933 |
+# Change permissions and set xattr on dir4
|
|
|
233933 |
+
|
|
|
233933 |
+TEST chmod 777 $M0/dir2
|
|
|
233933 |
+
|
|
|
233933 |
+TEST setfattr -n "user.test" -v "test" $M0/dir3
|
|
|
233933 |
+
|
|
|
233933 |
+TEST chmod 777 $M0/dir4
|
|
|
233933 |
+TEST setfattr -n "user.test" -v "test" $M0/dir4
|
|
|
233933 |
+
|
|
|
233933 |
+
|
|
|
233933 |
+# Start all bricks
|
|
|
233933 |
+
|
|
|
233933 |
+TEST $CLI volume start $V0 force
|
|
|
233933 |
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "1" brick_up_status $V0 $H0 $B0/$V0-1
|
|
|
233933 |
+
|
|
|
233933 |
+#$CLI volume status
|
|
|
233933 |
+
|
|
|
233933 |
+# It takes a while for the client to reconnect to the brick
|
|
|
233933 |
+sleep 5
|
|
|
233933 |
+
|
|
|
233933 |
+stat $M0/dir* > /dev/null
|
|
|
233933 |
+
|
|
|
233933 |
+# Check directories on the brick that was killed
|
|
|
233933 |
+
|
|
|
233933 |
+TEST ls $B0/$V0-1/dir2
|
|
|
233933 |
+EXPECT "777" stat -c "%a" $B0/$V0-1/dir2
|
|
|
233933 |
+
|
|
|
233933 |
+TEST ls $B0/$V0-1/dir3
|
|
|
233933 |
+EXPECT "test" getfattr -n "user.test" --absolute-names --only-values $B0/$V0-1/dir3
|
|
|
233933 |
+
|
|
|
233933 |
+
|
|
|
233933 |
+TEST ls $B0/$V0-1/dir4
|
|
|
233933 |
+EXPECT "777" stat -c "%a" $B0/$V0-1/dir4
|
|
|
233933 |
+EXPECT "test" getfattr -n "user.test" --absolute-names --only-values $B0/$V0-1/dir4
|
|
|
233933 |
+cd
|
|
|
233933 |
+
|
|
|
233933 |
+
|
|
|
233933 |
+# Cleanup
|
|
|
233933 |
+cleanup
|
|
|
233933 |
+
|
|
|
233933 |
diff --git a/xlators/cluster/dht/src/dht-common.c b/xlators/cluster/dht/src/dht-common.c
|
|
|
233933 |
index 2a68193..d3e900c 100644
|
|
|
233933 |
--- a/xlators/cluster/dht/src/dht-common.c
|
|
|
233933 |
+++ b/xlators/cluster/dht/src/dht-common.c
|
|
|
233933 |
@@ -801,9 +801,8 @@ dht_common_mark_mdsxattr(call_frame_t *frame, int *errst,
|
|
|
233933 |
call_frame_t *xattr_frame = NULL;
|
|
|
233933 |
gf_boolean_t vol_down = _gf_false;
|
|
|
233933 |
|
|
|
233933 |
- this = frame->this;
|
|
|
233933 |
-
|
|
|
233933 |
GF_VALIDATE_OR_GOTO("dht", frame, out);
|
|
|
233933 |
+ this = frame->this;
|
|
|
233933 |
GF_VALIDATE_OR_GOTO("dht", this, out);
|
|
|
233933 |
GF_VALIDATE_OR_GOTO(this->name, frame->local, out);
|
|
|
233933 |
GF_VALIDATE_OR_GOTO(this->name, this->private, out);
|
|
|
233933 |
@@ -812,6 +811,7 @@ dht_common_mark_mdsxattr(call_frame_t *frame, int *errst,
|
|
|
233933 |
conf = this->private;
|
|
|
233933 |
layout = local->selfheal.layout;
|
|
|
233933 |
local->mds_heal_fresh_lookup = mark_during_fresh_lookup;
|
|
|
233933 |
+
|
|
|
233933 |
gf_uuid_unparse(local->gfid, gfid_local);
|
|
|
233933 |
|
|
|
233933 |
/* Code to update hashed subvol consider as a mds subvol
|
|
|
233933 |
@@ -1240,6 +1240,31 @@ out:
|
|
|
233933 |
}
|
|
|
233933 |
|
|
|
233933 |
int
|
|
|
233933 |
+dht_needs_selfheal(call_frame_t *frame, xlator_t *this)
|
|
|
233933 |
+{
|
|
|
233933 |
+ dht_local_t *local = NULL;
|
|
|
233933 |
+ dht_layout_t *layout = NULL;
|
|
|
233933 |
+ int needs_selfheal = 0;
|
|
|
233933 |
+ int ret = 0;
|
|
|
233933 |
+
|
|
|
233933 |
+ local = frame->local;
|
|
|
233933 |
+ layout = local->layout;
|
|
|
233933 |
+
|
|
|
233933 |
+ if (local->need_attrheal || local->need_xattr_heal ||
|
|
|
233933 |
+ local->need_selfheal) {
|
|
|
233933 |
+ needs_selfheal = 1;
|
|
|
233933 |
+ }
|
|
|
233933 |
+
|
|
|
233933 |
+ ret = dht_layout_normalize(this, &local->loc, layout);
|
|
|
233933 |
+
|
|
|
233933 |
+ if (ret != 0) {
|
|
|
233933 |
+ gf_msg_debug(this->name, 0, "fixing assignment on %s", local->loc.path);
|
|
|
233933 |
+ needs_selfheal = 1;
|
|
|
233933 |
+ }
|
|
|
233933 |
+ return needs_selfheal;
|
|
|
233933 |
+}
|
|
|
233933 |
+
|
|
|
233933 |
+int
|
|
|
233933 |
dht_lookup_dir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
|
|
|
233933 |
int op_ret, int op_errno, inode_t *inode, struct iatt *stbuf,
|
|
|
233933 |
dict_t *xattr, struct iatt *postparent)
|
|
|
233933 |
@@ -1256,8 +1281,6 @@ dht_lookup_dir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
|
|
|
233933 |
char gfid_local[GF_UUID_BUF_SIZE] = {0};
|
|
|
233933 |
char gfid_node[GF_UUID_BUF_SIZE] = {0};
|
|
|
233933 |
int32_t mds_xattr_val[1] = {0};
|
|
|
233933 |
- call_frame_t *copy = NULL;
|
|
|
233933 |
- dht_local_t *copy_local = NULL;
|
|
|
233933 |
|
|
|
233933 |
GF_VALIDATE_OR_GOTO("dht", frame, out);
|
|
|
233933 |
GF_VALIDATE_OR_GOTO("dht", this, out);
|
|
|
233933 |
@@ -1270,7 +1293,11 @@ dht_lookup_dir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
|
|
|
233933 |
conf = this->private;
|
|
|
233933 |
|
|
|
233933 |
layout = local->layout;
|
|
|
233933 |
+ gf_msg_debug(this->name, op_errno,
|
|
|
233933 |
+ "%s: lookup on %s returned with op_ret = %d, op_errno = %d",
|
|
|
233933 |
+ local->loc.path, prev->name, op_ret, op_errno);
|
|
|
233933 |
|
|
|
233933 |
+ /* The first successful lookup*/
|
|
|
233933 |
if (!op_ret && gf_uuid_is_null(local->gfid)) {
|
|
|
233933 |
memcpy(local->gfid, stbuf->ia_gfid, 16);
|
|
|
233933 |
}
|
|
|
233933 |
@@ -1298,13 +1325,10 @@ dht_lookup_dir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
|
|
|
233933 |
|
|
|
233933 |
if (op_ret == -1) {
|
|
|
233933 |
local->op_errno = op_errno;
|
|
|
233933 |
- gf_msg_debug(this->name, op_errno,
|
|
|
233933 |
- "%s: lookup on %s returned error", local->loc.path,
|
|
|
233933 |
- prev->name);
|
|
|
233933 |
|
|
|
233933 |
/* The GFID is missing on this subvol. Force a heal. */
|
|
|
233933 |
if (op_errno == ENODATA) {
|
|
|
233933 |
- local->need_selfheal = 1;
|
|
|
233933 |
+ local->need_lookup_everywhere = 1;
|
|
|
233933 |
}
|
|
|
233933 |
goto unlock;
|
|
|
233933 |
}
|
|
|
233933 |
@@ -1312,12 +1336,11 @@ dht_lookup_dir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
|
|
|
233933 |
is_dir = check_is_dir(inode, stbuf, xattr);
|
|
|
233933 |
if (!is_dir) {
|
|
|
233933 |
gf_msg_debug(this->name, 0,
|
|
|
233933 |
- "lookup of %s on %s returned non"
|
|
|
233933 |
- "dir 0%o"
|
|
|
233933 |
+ "%s: lookup on %s returned non dir 0%o"
|
|
|
233933 |
"calling lookup_everywhere",
|
|
|
233933 |
local->loc.path, prev->name, stbuf->ia_type);
|
|
|
233933 |
|
|
|
233933 |
- local->need_selfheal = 1;
|
|
|
233933 |
+ local->need_lookup_everywhere = 1;
|
|
|
233933 |
goto unlock;
|
|
|
233933 |
}
|
|
|
233933 |
|
|
|
233933 |
@@ -1328,14 +1351,8 @@ dht_lookup_dir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
|
|
|
233933 |
dht_aggregate_xattr(local->xattr, xattr);
|
|
|
233933 |
}
|
|
|
233933 |
|
|
|
233933 |
- if (dict_get(xattr, conf->mds_xattr_key)) {
|
|
|
233933 |
- local->mds_subvol = prev;
|
|
|
233933 |
- local->mds_stbuf.ia_gid = stbuf->ia_gid;
|
|
|
233933 |
- local->mds_stbuf.ia_uid = stbuf->ia_uid;
|
|
|
233933 |
- local->mds_stbuf.ia_prot = stbuf->ia_prot;
|
|
|
233933 |
- }
|
|
|
233933 |
-
|
|
|
233933 |
if (local->stbuf.ia_type != IA_INVAL) {
|
|
|
233933 |
+ /* This is not the first subvol to respond */
|
|
|
233933 |
if (!__is_root_gfid(stbuf->ia_gfid) &&
|
|
|
233933 |
((local->stbuf.ia_gid != stbuf->ia_gid) ||
|
|
|
233933 |
(local->stbuf.ia_uid != stbuf->ia_uid) ||
|
|
|
233933 |
@@ -1348,65 +1365,64 @@ dht_lookup_dir_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
|
|
|
233933 |
if (local->inode == NULL)
|
|
|
233933 |
local->inode = inode_ref(inode);
|
|
|
233933 |
|
|
|
233933 |
+ /* This could be a problem */
|
|
|
233933 |
dht_iatt_merge(this, &local->stbuf, stbuf);
|
|
|
233933 |
dht_iatt_merge(this, &local->postparent, postparent);
|
|
|
233933 |
|
|
|
233933 |
if (!dict_get(xattr, conf->mds_xattr_key)) {
|
|
|
233933 |
gf_msg_debug(this->name, 0,
|
|
|
233933 |
- "Internal xattr %s is not present "
|
|
|
233933 |
- " on path %s gfid is %s ",
|
|
|
233933 |
- conf->mds_xattr_key, local->loc.path, gfid_local);
|
|
|
233933 |
+ "%s: mds xattr %s is not present "
|
|
|
233933 |
+ "on %s(gfid = %s)",
|
|
|
233933 |
+ local->loc.path, conf->mds_xattr_key, prev->name,
|
|
|
233933 |
+ gfid_local);
|
|
|
233933 |
goto unlock;
|
|
|
233933 |
- } else {
|
|
|
233933 |
- /* Save mds subvol on inode ctx */
|
|
|
233933 |
- ret = dht_inode_ctx_mdsvol_set(local->inode, this, prev);
|
|
|
233933 |
- if (ret) {
|
|
|
233933 |
- gf_msg(this->name, GF_LOG_ERROR, 0,
|
|
|
233933 |
- DHT_MSG_SET_INODE_CTX_FAILED,
|
|
|
233933 |
- "Failed to set hashed subvol for %s vol is %s",
|
|
|
233933 |
- local->loc.path, prev->name);
|
|
|
233933 |
- }
|
|
|
233933 |
+ }
|
|
|
233933 |
+
|
|
|
233933 |
+ local->mds_subvol = prev;
|
|
|
233933 |
+ local->mds_stbuf = *stbuf;
|
|
|
233933 |
+
|
|
|
233933 |
+ /* Save mds subvol on inode ctx */
|
|
|
233933 |
+
|
|
|
233933 |
+ ret = dht_inode_ctx_mdsvol_set(local->inode, this, prev);
|
|
|
233933 |
+ if (ret) {
|
|
|
233933 |
+ gf_msg(this->name, GF_LOG_ERROR, 0, DHT_MSG_SET_INODE_CTX_FAILED,
|
|
|
233933 |
+ "%s: Failed to set mds (%s)", local->loc.path, prev->name);
|
|
|
233933 |
}
|
|
|
233933 |
check_mds = dht_dict_get_array(xattr, conf->mds_xattr_key,
|
|
|
233933 |
mds_xattr_val, 1, &errst);
|
|
|
233933 |
if ((check_mds < 0) && !errst) {
|
|
|
233933 |
local->mds_xattr = dict_ref(xattr);
|
|
|
233933 |
gf_msg_debug(this->name, 0,
|
|
|
233933 |
- "Value of %s is not zero on hashed subvol "
|
|
|
233933 |
- "so xattr needs to be heal on non hashed"
|
|
|
233933 |
- " path is %s and vol name is %s "
|
|
|
233933 |
- " gfid is %s",
|
|
|
233933 |
- conf->mds_xattr_key, local->loc.path, prev->name,
|
|
|
233933 |
+ "%s: %s is not zero on %s. Xattrs need to be healed."
|
|
|
233933 |
+ "(gfid = %s)",
|
|
|
233933 |
+ local->loc.path, conf->mds_xattr_key, prev->name,
|
|
|
233933 |
gfid_local);
|
|
|
233933 |
local->need_xattr_heal = 1;
|
|
|
233933 |
- local->mds_subvol = prev;
|
|
|
233933 |
}
|
|
|
233933 |
}
|
|
|
233933 |
+
|
|
|
233933 |
unlock:
|
|
|
233933 |
UNLOCK(&frame->lock);
|
|
|
233933 |
|
|
|
233933 |
this_call_cnt = dht_frame_return(frame);
|
|
|
233933 |
|
|
|
233933 |
if (is_last_call(this_call_cnt)) {
|
|
|
233933 |
+ /* If the mds subvol is not set correctly*/
|
|
|
233933 |
+ if (!__is_root_gfid(local->gfid) &&
|
|
|
233933 |
+ (!dict_get(local->xattr, conf->mds_xattr_key))) {
|
|
|
233933 |
+ local->need_selfheal = 1;
|
|
|
233933 |
+ }
|
|
|
233933 |
+
|
|
|
233933 |
/* No need to call xattr heal code if volume count is 1
|
|
|
233933 |
*/
|
|
|
233933 |
- if (conf->subvolume_cnt == 1)
|
|
|
233933 |
+ if (conf->subvolume_cnt == 1) {
|
|
|
233933 |
local->need_xattr_heal = 0;
|
|
|
233933 |
-
|
|
|
233933 |
- /* Code to update all extended attributed from hashed subvol
|
|
|
233933 |
- to local->xattr
|
|
|
233933 |
- */
|
|
|
233933 |
- if (local->need_xattr_heal && (local->mds_xattr)) {
|
|
|
233933 |
- dht_dir_set_heal_xattr(this, local, local->xattr, local->mds_xattr,
|
|
|
233933 |
- NULL, NULL);
|
|
|
233933 |
- dict_unref(local->mds_xattr);
|
|
|
233933 |
- local->mds_xattr = NULL;
|
|
|
233933 |
}
|
|
|
233933 |
|
|
|
233933 |
- if (local->need_selfheal) {
|
|
|
233933 |
- local->need_selfheal = 0;
|
|
|
233933 |
+ if (local->need_selfheal || local->need_lookup_everywhere) {
|
|
|
233933 |
/* Set the gfid-req so posix will set the GFID*/
|
|
|
233933 |
if (!gf_uuid_is_null(local->gfid)) {
|
|
|
233933 |
+ /* Ok, this should _never_ happen */
|
|
|
233933 |
ret = dict_set_static_bin(local->xattr_req, "gfid-req",
|
|
|
233933 |
local->gfid, 16);
|
|
|
233933 |
} else {
|
|
|
233933 |
@@ -1414,73 +1430,36 @@ unlock:
|
|
|
233933 |
ret = dict_set_static_bin(local->xattr_req, "gfid-req",
|
|
|
233933 |
local->gfid_req, 16);
|
|
|
233933 |
}
|
|
|
233933 |
+ }
|
|
|
233933 |
+
|
|
|
233933 |
+ if (local->need_lookup_everywhere) {
|
|
|
233933 |
+ local->need_lookup_everywhere = 0;
|
|
|
233933 |
dht_lookup_everywhere(frame, this, &local->loc);
|
|
|
233933 |
return 0;
|
|
|
233933 |
}
|
|
|
233933 |
|
|
|
233933 |
if (local->op_ret == 0) {
|
|
|
233933 |
- ret = dht_layout_normalize(this, &local->loc, layout);
|
|
|
233933 |
-
|
|
|
233933 |
- if (ret != 0) {
|
|
|
233933 |
- gf_msg_debug(this->name, 0, "fixing assignment on %s",
|
|
|
233933 |
- local->loc.path);
|
|
|
233933 |
+ if (dht_needs_selfheal(frame, this)) {
|
|
|
233933 |
goto selfheal;
|
|
|
233933 |
}
|
|
|
233933 |
|
|
|
233933 |
dht_layout_set(this, local->inode, layout);
|
|
|
233933 |
- if (!dict_get(local->xattr, conf->mds_xattr_key) ||
|
|
|
233933 |
- local->need_xattr_heal)
|
|
|
233933 |
- goto selfheal;
|
|
|
233933 |
- }
|
|
|
233933 |
-
|
|
|
233933 |
- if (local->inode) {
|
|
|
233933 |
- dht_inode_ctx_time_update(local->inode, this, &local->stbuf, 1);
|
|
|
233933 |
- }
|
|
|
233933 |
-
|
|
|
233933 |
- if (local->loc.parent) {
|
|
|
233933 |
- dht_inode_ctx_time_update(local->loc.parent, this,
|
|
|
233933 |
- &local->postparent, 1);
|
|
|
233933 |
- }
|
|
|
233933 |
-
|
|
|
233933 |
- if (local->need_attrheal) {
|
|
|
233933 |
- local->need_attrheal = 0;
|
|
|
233933 |
- if (!__is_root_gfid(inode->gfid)) {
|
|
|
233933 |
- local->stbuf.ia_gid = local->mds_stbuf.ia_gid;
|
|
|
233933 |
- local->stbuf.ia_uid = local->mds_stbuf.ia_uid;
|
|
|
233933 |
- local->stbuf.ia_prot = local->mds_stbuf.ia_prot;
|
|
|
233933 |
+ if (local->inode) {
|
|
|
233933 |
+ dht_inode_ctx_time_update(local->inode, this, &local->stbuf, 1);
|
|
|
233933 |
}
|
|
|
233933 |
- copy = create_frame(this, this->ctx->pool);
|
|
|
233933 |
- if (copy) {
|
|
|
233933 |
- copy_local = dht_local_init(copy, &local->loc, NULL, 0);
|
|
|
233933 |
- if (!copy_local) {
|
|
|
233933 |
- DHT_STACK_DESTROY(copy);
|
|
|
233933 |
- goto skip_attr_heal;
|
|
|
233933 |
- }
|
|
|
233933 |
- copy_local->stbuf = local->stbuf;
|
|
|
233933 |
- gf_uuid_copy(copy_local->loc.gfid, local->stbuf.ia_gfid);
|
|
|
233933 |
- copy_local->mds_stbuf = local->mds_stbuf;
|
|
|
233933 |
- copy_local->mds_subvol = local->mds_subvol;
|
|
|
233933 |
- copy->local = copy_local;
|
|
|
233933 |
- FRAME_SU_DO(copy, dht_local_t);
|
|
|
233933 |
- ret = synctask_new(this->ctx->env, dht_dir_attr_heal,
|
|
|
233933 |
- dht_dir_attr_heal_done, copy, copy);
|
|
|
233933 |
- if (ret) {
|
|
|
233933 |
- gf_msg(this->name, GF_LOG_ERROR, ENOMEM,
|
|
|
233933 |
- DHT_MSG_DIR_ATTR_HEAL_FAILED,
|
|
|
233933 |
- "Synctask creation failed to heal attr "
|
|
|
233933 |
- "for path %s gfid %s ",
|
|
|
233933 |
- local->loc.path, local->gfid);
|
|
|
233933 |
- DHT_STACK_DESTROY(copy);
|
|
|
233933 |
- }
|
|
|
233933 |
+
|
|
|
233933 |
+ if (local->loc.parent) {
|
|
|
233933 |
+ dht_inode_ctx_time_update(local->loc.parent, this,
|
|
|
233933 |
+ &local->postparent, 1);
|
|
|
233933 |
}
|
|
|
233933 |
}
|
|
|
233933 |
|
|
|
233933 |
- skip_attr_heal:
|
|
|
233933 |
DHT_STRIP_PHASE1_FLAGS(&local->stbuf);
|
|
|
233933 |
dht_set_fixed_dir_stat(&local->postparent);
|
|
|
233933 |
/* Delete mds xattr at the time of STACK UNWIND */
|
|
|
233933 |
if (local->xattr)
|
|
|
233933 |
GF_REMOVE_INTERNAL_XATTR(conf->mds_xattr_key, local->xattr);
|
|
|
233933 |
+
|
|
|
233933 |
DHT_STACK_UNWIND(lookup, frame, local->op_ret, local->op_errno,
|
|
|
233933 |
local->inode, &local->stbuf, local->xattr,
|
|
|
233933 |
&local->postparent);
|
|
|
233933 |
@@ -5444,9 +5423,8 @@ dht_dir_common_set_remove_xattr(call_frame_t *frame, xlator_t *this, loc_t *loc,
|
|
|
233933 |
} else {
|
|
|
233933 |
gf_msg(this->name, GF_LOG_ERROR, 0,
|
|
|
233933 |
DHT_MSG_HASHED_SUBVOL_GET_FAILED,
|
|
|
233933 |
- "Failed to get mds subvol for path %s"
|
|
|
233933 |
- "gfid is %s ",
|
|
|
233933 |
- loc->path, gfid_local);
|
|
|
233933 |
+ "%s: Failed to get mds subvol. (gfid is %s)", loc->path,
|
|
|
233933 |
+ gfid_local);
|
|
|
233933 |
}
|
|
|
233933 |
(*op_errno) = ENOENT;
|
|
|
233933 |
goto err;
|
|
|
233933 |
diff --git a/xlators/cluster/dht/src/dht-selfheal.c b/xlators/cluster/dht/src/dht-selfheal.c
|
|
|
233933 |
index bd1b7ea..5420fca 100644
|
|
|
233933 |
--- a/xlators/cluster/dht/src/dht-selfheal.c
|
|
|
233933 |
+++ b/xlators/cluster/dht/src/dht-selfheal.c
|
|
|
233933 |
@@ -1033,18 +1033,27 @@ dht_selfheal_dir_setattr(call_frame_t *frame, loc_t *loc, struct iatt *stbuf,
|
|
|
233933 |
int missing_attr = 0;
|
|
|
233933 |
int i = 0, ret = -1;
|
|
|
233933 |
dht_local_t *local = NULL;
|
|
|
233933 |
+ dht_conf_t *conf = NULL;
|
|
|
233933 |
xlator_t *this = NULL;
|
|
|
233933 |
int cnt = 0;
|
|
|
233933 |
|
|
|
233933 |
local = frame->local;
|
|
|
233933 |
this = frame->this;
|
|
|
233933 |
+ conf = this->private;
|
|
|
233933 |
+
|
|
|
233933 |
+ /* We need to heal the attrs if:
|
|
|
233933 |
+ * 1. Any directories were missing - the newly created dirs will need
|
|
|
233933 |
+ * to have the correct attrs set
|
|
|
233933 |
+ * 2. An existing dir does not have the correct permissions -they may
|
|
|
233933 |
+ * have been changed when a brick was down.
|
|
|
233933 |
+ */
|
|
|
233933 |
|
|
|
233933 |
for (i = 0; i < layout->cnt; i++) {
|
|
|
233933 |
if (layout->list[i].err == -1)
|
|
|
233933 |
missing_attr++;
|
|
|
233933 |
}
|
|
|
233933 |
|
|
|
233933 |
- if (missing_attr == 0) {
|
|
|
233933 |
+ if ((missing_attr == 0) && (local->need_attrheal == 0)) {
|
|
|
233933 |
if (!local->heal_layout) {
|
|
|
233933 |
gf_msg_trace(this->name, 0, "Skip heal layout for %s gfid = %s ",
|
|
|
233933 |
loc->path, uuid_utoa(loc->gfid));
|
|
|
233933 |
@@ -1062,19 +1071,12 @@ dht_selfheal_dir_setattr(call_frame_t *frame, loc_t *loc, struct iatt *stbuf,
|
|
|
233933 |
return 0;
|
|
|
233933 |
}
|
|
|
233933 |
|
|
|
233933 |
- local->call_cnt = missing_attr;
|
|
|
233933 |
- cnt = layout->cnt;
|
|
|
233933 |
+ cnt = local->call_cnt = conf->subvolume_cnt;
|
|
|
233933 |
|
|
|
233933 |
for (i = 0; i < cnt; i++) {
|
|
|
233933 |
- if (layout->list[i].err == -1) {
|
|
|
233933 |
- gf_msg_trace(this->name, 0, "%s: setattr on subvol %s, gfid = %s",
|
|
|
233933 |
- loc->path, layout->list[i].xlator->name,
|
|
|
233933 |
- uuid_utoa(loc->gfid));
|
|
|
233933 |
-
|
|
|
233933 |
- STACK_WIND(
|
|
|
233933 |
- frame, dht_selfheal_dir_setattr_cbk, layout->list[i].xlator,
|
|
|
233933 |
- layout->list[i].xlator->fops->setattr, loc, stbuf, valid, NULL);
|
|
|
233933 |
- }
|
|
|
233933 |
+ STACK_WIND(frame, dht_selfheal_dir_setattr_cbk, layout->list[i].xlator,
|
|
|
233933 |
+ layout->list[i].xlator->fops->setattr, loc, stbuf, valid,
|
|
|
233933 |
+ NULL);
|
|
|
233933 |
}
|
|
|
233933 |
|
|
|
233933 |
return 0;
|
|
|
233933 |
@@ -1492,6 +1494,9 @@ dht_selfheal_dir_mkdir(call_frame_t *frame, loc_t *loc, dht_layout_t *layout,
|
|
|
233933 |
}
|
|
|
233933 |
|
|
|
233933 |
if (missing_dirs == 0) {
|
|
|
233933 |
+ /* We don't need to create any directories. Proceed to heal the
|
|
|
233933 |
+ * attrs and xattrs
|
|
|
233933 |
+ */
|
|
|
233933 |
if (!__is_root_gfid(local->stbuf.ia_gfid)) {
|
|
|
233933 |
if (local->need_xattr_heal) {
|
|
|
233933 |
local->need_xattr_heal = 0;
|
|
|
233933 |
@@ -1499,8 +1504,8 @@ dht_selfheal_dir_mkdir(call_frame_t *frame, loc_t *loc, dht_layout_t *layout,
|
|
|
233933 |
if (ret)
|
|
|
233933 |
gf_msg(this->name, GF_LOG_ERROR, ret,
|
|
|
233933 |
DHT_MSG_DIR_XATTR_HEAL_FAILED,
|
|
|
233933 |
- "xattr heal failed for "
|
|
|
233933 |
- "directory %s gfid %s ",
|
|
|
233933 |
+ "%s:xattr heal failed for "
|
|
|
233933 |
+ "directory (gfid = %s)",
|
|
|
233933 |
local->loc.path, local->gfid);
|
|
|
233933 |
} else {
|
|
|
233933 |
if (!gf_uuid_is_null(local->gfid))
|
|
|
233933 |
@@ -1512,8 +1517,8 @@ dht_selfheal_dir_mkdir(call_frame_t *frame, loc_t *loc, dht_layout_t *layout,
|
|
|
233933 |
|
|
|
233933 |
gf_msg(this->name, GF_LOG_INFO, 0,
|
|
|
233933 |
DHT_MSG_DIR_XATTR_HEAL_FAILED,
|
|
|
233933 |
- "Failed to set mds xattr "
|
|
|
233933 |
- "for directory %s gfid %s ",
|
|
|
233933 |
+ "%s: Failed to set mds xattr "
|
|
|
233933 |
+ "for directory (gfid = %s)",
|
|
|
233933 |
local->loc.path, local->gfid);
|
|
|
233933 |
}
|
|
|
233933 |
}
|
|
|
233933 |
@@ -2085,10 +2090,10 @@ dht_selfheal_directory(call_frame_t *frame, dht_selfheal_dir_cbk_t dir_cbk,
|
|
|
233933 |
loc_t *loc, dht_layout_t *layout)
|
|
|
233933 |
{
|
|
|
233933 |
dht_local_t *local = NULL;
|
|
|
233933 |
+ xlator_t *this = NULL;
|
|
|
233933 |
uint32_t down = 0;
|
|
|
233933 |
uint32_t misc = 0;
|
|
|
233933 |
int ret = 0;
|
|
|
233933 |
- xlator_t *this = NULL;
|
|
|
233933 |
char pgfid[GF_UUID_BUF_SIZE] = {0};
|
|
|
233933 |
char gfid[GF_UUID_BUF_SIZE] = {0};
|
|
|
233933 |
inode_t *linked_inode = NULL, *inode = NULL;
|
|
|
233933 |
@@ -2099,6 +2104,11 @@ dht_selfheal_directory(call_frame_t *frame, dht_selfheal_dir_cbk_t dir_cbk,
|
|
|
233933 |
local->selfheal.dir_cbk = dir_cbk;
|
|
|
233933 |
local->selfheal.layout = dht_layout_ref(this, layout);
|
|
|
233933 |
|
|
|
233933 |
+ if (local->need_attrheal && !IA_ISINVAL(local->mds_stbuf.ia_type)) {
|
|
|
233933 |
+ /*Use the one in the mds_stbuf*/
|
|
|
233933 |
+ local->stbuf = local->mds_stbuf;
|
|
|
233933 |
+ }
|
|
|
233933 |
+
|
|
|
233933 |
if (!__is_root_gfid(local->stbuf.ia_gfid)) {
|
|
|
233933 |
gf_uuid_unparse(local->stbuf.ia_gfid, gfid);
|
|
|
233933 |
gf_uuid_unparse(loc->parent->gfid, pgfid);
|
|
|
233933 |
@@ -2118,6 +2128,13 @@ dht_selfheal_directory(call_frame_t *frame, dht_selfheal_dir_cbk_t dir_cbk,
|
|
|
233933 |
inode_unref(inode);
|
|
|
233933 |
}
|
|
|
233933 |
|
|
|
233933 |
+ if (local->need_xattr_heal && (local->mds_xattr)) {
|
|
|
233933 |
+ dht_dir_set_heal_xattr(this, local, local->xattr, local->mds_xattr,
|
|
|
233933 |
+ NULL, NULL);
|
|
|
233933 |
+ dict_unref(local->mds_xattr);
|
|
|
233933 |
+ local->mds_xattr = NULL;
|
|
|
233933 |
+ }
|
|
|
233933 |
+
|
|
|
233933 |
dht_layout_anomalies(this, loc, layout, &local->selfheal.hole_cnt,
|
|
|
233933 |
&local->selfheal.overlaps_cnt,
|
|
|
233933 |
&local->selfheal.missing_cnt, &local->selfheal.down,
|
|
|
233933 |
@@ -2128,18 +2145,18 @@ dht_selfheal_directory(call_frame_t *frame, dht_selfheal_dir_cbk_t dir_cbk,
|
|
|
233933 |
|
|
|
233933 |
if (down) {
|
|
|
233933 |
gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_DIR_SELFHEAL_FAILED,
|
|
|
233933 |
- "Directory selfheal failed: %d subvolumes down."
|
|
|
233933 |
- "Not fixing. path = %s, gfid = %s",
|
|
|
233933 |
- down, loc->path, gfid);
|
|
|
233933 |
+ "%s: Directory selfheal failed: %d subvolumes down."
|
|
|
233933 |
+ "Not fixing. gfid = %s",
|
|
|
233933 |
+ loc->path, down, gfid);
|
|
|
233933 |
ret = 0;
|
|
|
233933 |
goto sorry_no_fix;
|
|
|
233933 |
}
|
|
|
233933 |
|
|
|
233933 |
if (misc) {
|
|
|
233933 |
gf_msg(this->name, GF_LOG_WARNING, 0, DHT_MSG_DIR_SELFHEAL_FAILED,
|
|
|
233933 |
- "Directory selfheal failed : %d subvolumes "
|
|
|
233933 |
- "have unrecoverable errors. path = %s, gfid = %s",
|
|
|
233933 |
- misc, loc->path, gfid);
|
|
|
233933 |
+ "%s: Directory selfheal failed : %d subvolumes "
|
|
|
233933 |
+ "have unrecoverable errors. gfid = %s",
|
|
|
233933 |
+ loc->path, misc, gfid);
|
|
|
233933 |
|
|
|
233933 |
ret = 0;
|
|
|
233933 |
goto sorry_no_fix;
|
|
|
233933 |
@@ -2369,13 +2386,13 @@ dht_dir_attr_heal(void *data)
|
|
|
233933 |
|
|
|
233933 |
frame = data;
|
|
|
233933 |
local = frame->local;
|
|
|
233933 |
- mds_subvol = local->mds_subvol;
|
|
|
233933 |
this = frame->this;
|
|
|
233933 |
GF_VALIDATE_OR_GOTO("dht", this, out);
|
|
|
233933 |
GF_VALIDATE_OR_GOTO("dht", local, out);
|
|
|
233933 |
conf = this->private;
|
|
|
233933 |
GF_VALIDATE_OR_GOTO("dht", conf, out);
|
|
|
233933 |
|
|
|
233933 |
+ mds_subvol = local->mds_subvol;
|
|
|
233933 |
call_cnt = conf->subvolume_cnt;
|
|
|
233933 |
|
|
|
233933 |
if (!__is_root_gfid(local->stbuf.ia_gfid) && (!mds_subvol)) {
|
|
|
233933 |
--
|
|
|
233933 |
1.8.3.1
|
|
|
233933 |
|