summaryrefslogtreecommitdiffstats
path: root/xlators/cluster/dht/src/tier.c
diff options
context:
space:
mode:
authorMilind Changire <mchangir@redhat.com>2016-10-15 10:49:19 +0530
committerDan Lambright <dlambrig@redhat.com>2016-10-19 12:51:48 -0700
commit460016428cf27484c333227f534c2e2f73a37fb1 (patch)
tree321a1bb0517f622a134b3ab0634d6fea842fdb06 /xlators/cluster/dht/src/tier.c
parent20d2b36f3acc2f27527b7913d2ad939848395aeb (diff)
cluster/tier: handle fast demotions
Demote files on priority if hi-watermark has been breached and continue to demote until the watermark drops below hi-watermark. Monitor watermark more frequently. Trigger demotion as soon as hi-watermark is breached. Add cluster.tier-emergency-demote-query-limit option to limit number of files returned from the database query for every iteration of tier_migrate_using_query_file(). If watermark hasn't dropped below hi-watermark during the first iteration, the next iteration will be triggered approximately 1 second after tier_demote() returns to the main tiering loop. Update changetimerecorder xlator to handle query for emergency demote mode. Add tier-ctr-interface.h: Move tier and ctr interface specific macros and struct definition from libglusterfs/src/gfdb/gfdb_data_store.h to new header libglusterfs/src/tier-ctr-interface.h Change-Id: If56af78c6c81d37529b9b6e65ae606ba5c99a811 BUG: 1366648 Signed-off-by: Milind Changire <mchangir@redhat.com> Reviewed-on: http://review.gluster.org/15158 Smoke: Gluster Build System <jenkins@build.gluster.org> CentOS-regression: Gluster Build System <jenkins@build.gluster.org> NetBSD-regression: NetBSD Build System <jenkins@build.gluster.org> Reviewed-by: Dan Lambright <dlambrig@redhat.com>
Diffstat (limited to 'xlators/cluster/dht/src/tier.c')
-rw-r--r--xlators/cluster/dht/src/tier.c93
1 files changed, 78 insertions, 15 deletions
diff --git a/xlators/cluster/dht/src/tier.c b/xlators/cluster/dht/src/tier.c
index 2f8eddbc07d..60e967a88a2 100644
--- a/xlators/cluster/dht/src/tier.c
+++ b/xlators/cluster/dht/src/tier.c
@@ -15,6 +15,7 @@
#include "tier-common.h"
#include "syscall.h"
#include "events.h"
+#include "tier-ctr-interface.h"
/*Hard coded DB info*/
static gfdb_db_type_t dht_tier_db_type = GFDB_SQLITE3;
@@ -193,6 +194,7 @@ out:
/* Check and update the watermark every WM_INTERVAL seconds */
#define WM_INTERVAL 5
+#define WM_INTERVAL_EMERG 1
static int
tier_check_same_node (xlator_t *this, loc_t *loc, gf_defrag_info_t *defrag)
@@ -523,7 +525,7 @@ tier_can_promote_file (xlator_t *this, char const *file_name,
defrag->tier_conf.blocks_used;
/* test if the estimated block usage goes above HI watermark */
- if (GF_PERCENTAGE (estimated_usage, defrag->tier_conf.blocks_total) >
+ if (GF_PERCENTAGE (estimated_usage, defrag->tier_conf.blocks_total) >=
defrag->tier_conf.watermark_hi) {
gf_msg (this->name, GF_LOG_INFO, 0,
DHT_MSG_LOG_TIER_STATUS,
@@ -575,6 +577,7 @@ tier_migrate_using_query_file (void *_args)
gfdb_time_t current_time = { 0 };
int total_time = 0;
int max_time = 0;
+ gf_boolean_t emergency_demote_mode = _gf_false;
GF_VALIDATE_OR_GOTO ("tier", query_cbk_args, out);
@@ -591,6 +594,9 @@ tier_migrate_using_query_file (void *_args)
if (!migrate_data)
goto out;
+ emergency_demote_mode = (!query_cbk_args->is_promotion &&
+ is_hot_tier_full(&defrag->tier_conf));
+
xdata_request = dict_new ();
if (!xdata_request) {
gf_msg (this->name, GF_LOG_ERROR, 0,
@@ -1013,6 +1019,18 @@ per_file_out:
gfdb_methods.gfdb_query_record_free (query_record);
query_record = NULL;
+
+ /* If we are demoting and the entry watermark was HI, then
+ * we are done with emergency demotions if the current
+ * watermark has fallen below hi-watermark level
+ */
+ if (emergency_demote_mode) {
+ if (tier_check_watermark (this) == 0) {
+ if (!is_hot_tier_full (&defrag->tier_conf)) {
+ break;
+ }
+ }
+ }
}
out:
@@ -1126,14 +1144,23 @@ tier_process_self_query (tier_brick_list_t *local_brick, void *args)
goto out;
}
if (!gfdb_brick_info->_gfdb_promote) {
- if (query_cbk_args->defrag->write_freq_threshold == 0 &&
- query_cbk_args->defrag->read_freq_threshold == 0) {
- ret = gfdb_methods.find_unchanged_for_time (
- conn_node,
- tier_gf_query_callback,
- (void *)query_cbk_args,
- gfdb_brick_info->time_stamp);
+ if (query_cbk_args->defrag->tier_conf.watermark_last ==
+ TIER_WM_HI) {
+ /* emergency demotion mode */
+ ret = gfdb_methods.find_all (conn_node,
+ tier_gf_query_callback,
+ (void *)query_cbk_args,
+ query_cbk_args->defrag->tier_conf.
+ query_limit);
} else {
+ if (query_cbk_args->defrag->write_freq_threshold == 0 &&
+ query_cbk_args->defrag->read_freq_threshold == 0) {
+ ret = gfdb_methods.find_unchanged_for_time (
+ conn_node,
+ tier_gf_query_callback,
+ (void *)query_cbk_args,
+ gfdb_brick_info->time_stamp);
+ } else {
ret = gfdb_methods.find_unchanged_for_time_freq (
conn_node,
tier_gf_query_callback,
@@ -1144,6 +1171,7 @@ tier_process_self_query (tier_brick_list_t *local_brick, void *args)
query_cbk_args->defrag->
read_freq_threshold,
_gf_false);
+ }
}
} else {
if (query_cbk_args->defrag->write_freq_threshold == 0 &&
@@ -1159,8 +1187,7 @@ tier_process_self_query (tier_brick_list_t *local_brick, void *args)
tier_gf_query_callback,
(void *)query_cbk_args,
gfdb_brick_info->time_stamp,
- query_cbk_args->defrag->
- write_freq_threshold,
+ query_cbk_args->defrag->write_freq_threshold,
query_cbk_args->defrag->read_freq_threshold,
_gf_false);
}
@@ -1267,10 +1294,21 @@ tier_process_ctr_query (tier_brick_list_t *local_brick, void *args)
/* set all the query params*/
ipc_ctr_params->is_promote = gfdb_brick_info->_gfdb_promote;
- ipc_ctr_params->write_freq_threshold = query_cbk_args->
- defrag->write_freq_threshold;
- ipc_ctr_params->read_freq_threshold = query_cbk_args->
- defrag->read_freq_threshold;
+
+ ipc_ctr_params->write_freq_threshold =
+ query_cbk_args->defrag->write_freq_threshold;
+
+ ipc_ctr_params->read_freq_threshold =
+ query_cbk_args->defrag->read_freq_threshold;
+
+ ipc_ctr_params->query_limit =
+ query_cbk_args->defrag->tier_conf.query_limit;
+
+ ipc_ctr_params->emergency_demote =
+ (!gfdb_brick_info->_gfdb_promote &&
+ query_cbk_args->defrag->tier_conf.watermark_last ==
+ TIER_WM_HI);
+
memcpy (&ipc_ctr_params->time_stamp,
gfdb_brick_info->time_stamp,
sizeof (gfdb_time_t));
@@ -2208,6 +2246,15 @@ out:
return ret;
}
+static int
+tier_get_wm_interval(tier_mode_t mode, tier_watermark_op_t wm)
+{
+ if (mode == TIER_MODE_WM && wm == TIER_WM_HI)
+ return WM_INTERVAL_EMERG;
+
+ return WM_INTERVAL;
+}
+
/*
* Main tiering loop. This is called from the promotion and the
* demotion threads spawned in tier_start().
@@ -2316,7 +2363,10 @@ static void
check_watermark++;
- if (check_watermark >= WM_INTERVAL) {
+ /* emergency demotion requires frequent watermark monitoring */
+ if (check_watermark >=
+ tier_get_wm_interval(tier_conf->mode,
+ tier_conf->watermark_last)) {
check_watermark = 0;
if (tier_conf->mode == TIER_MODE_WM) {
ret = tier_get_fs_stat (this, &root_loc);
@@ -2828,6 +2878,15 @@ tier_init (xlator_t *this)
defrag->tier_conf.max_migrate_files = freq;
+
+ ret = dict_get_int32 (this->options,
+ "tier-query-limit",
+ &(defrag->tier_conf.query_limit));
+ if (ret) {
+ defrag->tier_conf.query_limit =
+ DEFAULT_TIER_QUERY_LIMIT;
+ }
+
ret = dict_get_str (this->options,
"tier-compact", &mode);
@@ -3041,6 +3100,10 @@ tier_reconfigure (xlator_t *this, dict_t *options)
defrag->tier_conf.max_migrate_files, options,
int32, out);
+ GF_OPTION_RECONF ("tier-query-limit",
+ defrag->tier_conf.query_limit,
+ options, int32, out);
+
GF_OPTION_RECONF ("tier-pause",
req_pause, options,
bool, out);