summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorAshish Pandey <aspandey@redhat.com>2017-01-11 17:19:30 +0530
committerPranith Kumar Karampuri <pkarampu@redhat.com>2017-01-20 04:29:34 -0800
commit578e9b5b5b45245ed044bab066533411e2141db6 (patch)
tree6ae1b4da7beef383d87cb985e14c094586a5b584
parent07b9853ad0c92b341be33a6cd632013c416221c8 (diff)
cluster/ec: Do not start heal on good file while IO is going on
Problem: Write on a file has been slowed down significantly after http://review.gluster.org/#/c/13733/ RC : When update fop starts on a file, it sets dirty flag at the start and remove it at the end which make an index entry in indices/xattrop. During IO, SHD scans this and finds out an index and starts heal even if all the fragments are healthy and up tp date. This heal takes inodelk for different types of heal. If the IO is for long time this will happen in every 60 seconds. Due to this extra, unneccessary locking, IO gets slowed down. Solution: Before starting any type of heal check if file needs heal or not. Change-Id: Ib9519a43e7e4b2565d3f3153f9ca0fb92174fe51 BUG: 1409191 Signed-off-by: Ashish Pandey <aspandey@redhat.com> Reviewed-on: http://review.gluster.org/16377 NetBSD-regression: NetBSD Build System <jenkins@build.gluster.org> CentOS-regression: Gluster Build System <jenkins@build.gluster.org> Smoke: Gluster Build System <jenkins@build.gluster.org> Reviewed-by: Pranith Kumar Karampuri <pkarampu@redhat.com> Reviewed-by: Xavier Hernandez <xhernandez@datalab.es>
-rw-r--r--xlators/cluster/ec/src/ec-common.h4
-rw-r--r--xlators/cluster/ec/src/ec-heal.c41
2 files changed, 41 insertions, 4 deletions
diff --git a/xlators/cluster/ec/src/ec-common.h b/xlators/cluster/ec/src/ec-common.h
index e3fc3cef075..c532e0e03b0 100644
--- a/xlators/cluster/ec/src/ec-common.h
+++ b/xlators/cluster/ec/src/ec-common.h
@@ -119,6 +119,10 @@ gf_boolean_t ec_is_recoverable_error (int32_t op_errno);
void ec_handle_healers_done (ec_fop_data_t *fop);
int32_t
+ec_heal_inspect (call_frame_t *frame, ec_t *ec,
+ inode_t *inode, unsigned char *locked_on,
+ gf_boolean_t *need_heal);
+int32_t
ec_get_heal_info (xlator_t *this, loc_t *loc, dict_t **dict);
#endif /* __EC_COMMON_H__ */
diff --git a/xlators/cluster/ec/src/ec-heal.c b/xlators/cluster/ec/src/ec-heal.c
index 1859d73d12b..548649b9a35 100644
--- a/xlators/cluster/ec/src/ec-heal.c
+++ b/xlators/cluster/ec/src/ec-heal.c
@@ -2332,6 +2332,8 @@ ec_heal_do (xlator_t *this, void *data, loc_t *loc, int32_t partial)
intptr_t bad = 0;
ec_fop_data_t *fop = data;
gf_boolean_t blocking = _gf_false;
+ gf_boolean_t need_heal = _gf_false;
+ unsigned char *up_subvols = NULL;
ec = this->private;
@@ -2353,6 +2355,19 @@ ec_heal_do (xlator_t *this, void *data, loc_t *loc, int32_t partial)
frame->root->pid = GF_CLIENT_PID_SELF_HEALD;
participants = alloca0(ec->nodes);
ec_mask_to_char_array (ec->xl_up, participants, ec->nodes);
+
+ up_subvols = alloca0(ec->nodes);
+ ec_mask_to_char_array (ec->xl_up, up_subvols, ec->nodes);
+
+ ec_heal_inspect (frame, ec, loc->inode, up_subvols,
+ &need_heal);
+ if (!need_heal) {
+ gf_msg (ec->xl->name, GF_LOG_DEBUG, 0,
+ EC_MSG_HEAL_FAIL, "Heal is not required for : %s ",
+ uuid_utoa(loc->gfid));
+ goto out;
+ }
+
if (loc->name && strlen (loc->name)) {
ret = ec_heal_name (frame, ec, loc->parent, (char *)loc->name,
participants);
@@ -2401,7 +2416,7 @@ ec_heal_do (xlator_t *this, void *data, loc_t *loc, int32_t partial)
op_errno = -ret;
}
-
+out:
if (fop->cbks.heal) {
fop->cbks.heal (fop->req_frame, fop, fop->xl, op_ret,
op_errno, ec_char_array_to_mask (participants,
@@ -2665,7 +2680,8 @@ out:
}
int32_t
-ec_need_heal (ec_t *ec, default_args_cbk_t *replies, gf_boolean_t *need_heal)
+ec_need_heal (ec_t *ec, default_args_cbk_t *replies,
+ gf_boolean_t *need_heal, int32_t lock_count)
{
uint64_t *dirty = NULL;
unsigned char *sources = NULL;
@@ -2691,7 +2707,9 @@ ec_need_heal (ec_t *ec, default_args_cbk_t *replies, gf_boolean_t *need_heal)
goto out;
}
source_count = EC_COUNT (sources, ec->nodes);
- if (source_count != ec->nodes) {
+ if (source_count == ec->nodes && lock_count > 0) {
+ *need_heal = _gf_false;
+ } else {
*need_heal = _gf_true;
}
ret = source_count;
@@ -2705,12 +2723,14 @@ ec_heal_inspect (call_frame_t *frame, ec_t *ec,
gf_boolean_t *need_heal)
{
loc_t loc = {0};
+ int i = 0;
int ret = 0;
dict_t *xdata = NULL;
uint64_t zero_array[2] = {0};
uint64_t zero_value = 0;
unsigned char *output = NULL;
default_args_cbk_t *replies = NULL;
+ int32_t lock_count = 0;
EC_REPLIES_ALLOC (replies, ec->nodes);
output = alloca0 (ec->nodes);
@@ -2720,6 +2740,8 @@ ec_heal_inspect (call_frame_t *frame, ec_t *ec,
xdata = dict_new ();
if (!xdata ||
+ dict_set_str(xdata, GLUSTERFS_INODELK_DOM_COUNT,
+ ec->xl->name) ||
dict_set_static_bin (xdata, EC_XATTR_VERSION, zero_array,
sizeof (zero_array)) ||
dict_set_static_bin (xdata, EC_XATTR_DIRTY, zero_array,
@@ -2731,12 +2753,23 @@ ec_heal_inspect (call_frame_t *frame, ec_t *ec,
}
ret = cluster_lookup (ec->xl_list, locked_on, ec->nodes, replies,
output, frame, ec->xl, &loc, xdata);
+
if (ret != ec->nodes) {
ret = ec->nodes;
*need_heal = _gf_true;
goto out;
}
- ret = ec_need_heal (ec, replies, need_heal);
+
+ for (i = 0; i < ec->nodes; i++) {
+ if (!output[i] || !replies[i].xdata) {
+ continue;
+ }
+ if ((dict_get_int32 (replies[i].xdata, GLUSTERFS_INODELK_COUNT,
+ &lock_count) == 0) && lock_count > 0) {
+ break;
+ }
+ }
+ ret = ec_need_heal (ec, replies, need_heal, lock_count);
out:
cluster_replies_wipe (replies, ec->nodes);