cluster/ec: Make background healing optional behavior

Provide options to control number of active background heal count and qlen.

Change-Id: Idc2419219d881f47e7d2e9bbc1dcdd999b372033
BUG: 1237381
Signed-off-by: Pranith Kumar K <pkarampu@redhat.com>
Reviewed-on: http://review.gluster.org/11473
Reviewed-by: Xavier Hernandez <xhernandez@datalab.es>
Tested-by: Gluster Build System <jenkins@build.gluster.com>

3 files changed, 55 insertions, 11 deletions

diff --git a/xlators/cluster/ec/src/ec-heal.c b/xlators/cluster/ec/src/ec-heal.c
index 31685356db0..6ee1f9ee832 100644
--- a/xlators/cluster/ec/src/ec-heal.c
+++ b/xlators/cluster/ec/src/ec-heal.c
@@ -26,9 +26,6 @@
 #include "syncop-utils.h"
 #include "cluster-syncop.h"
 
-#define EC_MAX_BACKGROUND_HEALS 8
-#define EC_MAX_HEAL_WAITERS 128
-
 #define alloca0(size) ({void *__ptr; __ptr = alloca(size); memset(__ptr, 0, size); __ptr; })
 #define EC_COUNT(array, max) ({int __i; int __res = 0; for (__i = 0; __i < max; __i++) if (array[__i]) __res++; __res; })
 #define EC_INTERSECT(dst, src1, src2, max) ({int __i; for (__i = 0; __i < max; __i++) dst[__i] = src1[__i] && src2[__i]; })
@@ -2329,10 +2326,9 @@ __ec_dequeue_heals (ec_t *ec)
         if (list_empty (&ec->heal_waiting))
                 goto none;
 
-        if (ec->healers == EC_MAX_BACKGROUND_HEALS)
+        if ((ec->background_heals > 0) && (ec->healers >= ec->background_heals))
                 goto none;
 
-        GF_ASSERT (ec->healers < EC_MAX_BACKGROUND_HEALS);
         fop = list_entry(ec->heal_waiting.next, ec_fop_data_t, healer);
         ec->heal_waiters--;
         list_del_init(&fop->healer);
@@ -2400,12 +2396,14 @@ ec_heal_throttle (xlator_t *this, ec_fop_data_t *fop)
 
                 LOCK (&ec->lock);
                 {
-                        if (ec->heal_waiters >= EC_MAX_HEAL_WAITERS) {
-                                can_heal = _gf_false;
-                        } else {
+                        if ((ec->background_heals > 0) &&
+                            (ec->heal_wait_qlen + ec->background_heals) >
+                                             (ec->heal_waiters + ec->healers)) {
                                 list_add_tail(&fop->healer, &ec->heal_waiting);
                                 ec->heal_waiters++;
                                 fop = __ec_dequeue_heals (ec);
+                        } else {
+                                can_heal = _gf_false;
                         }
                 }
                 UNLOCK (&ec->lock);
diff --git a/xlators/cluster/ec/src/ec.c b/xlators/cluster/ec/src/ec.c
index dd51630ea79..e28f402e6fe 100644
--- a/xlators/cluster/ec/src/ec.c
+++ b/xlators/cluster/ec/src/ec.c
@@ -219,15 +219,35 @@ int32_t mem_acct_init(xlator_t * this)
     return 0;
 }
 
+void
+ec_configure_background_heal_opts (ec_t *ec, int background_heals,
+                                   int heal_wait_qlen)
+{
+        if (background_heals == 0) {
+                ec->heal_wait_qlen = 0;
+        } else {
+                ec->heal_wait_qlen = heal_wait_qlen;
+        }
+        ec->background_heals = background_heals;
+}
+
 int32_t
 reconfigure (xlator_t *this, dict_t *options)
 {
-        ec_t *ec = this->private;
+        ec_t     *ec              = this->private;
+        uint32_t heal_wait_qlen   = 0;
+        uint32_t background_heals = 0;
 
-        GF_OPTION_RECONF ("self-heal-daemon", ec->shd.enabled, options, bool, failed);
+        GF_OPTION_RECONF ("self-heal-daemon", ec->shd.enabled, options, bool,
+                          failed);
         GF_OPTION_RECONF ("iam-self-heal-daemon", ec->shd.iamshd, options,
                           bool, failed);
-
+        GF_OPTION_RECONF ("background-heals", background_heals, options,
+                          uint32, failed);
+        GF_OPTION_RECONF ("heal-wait-qlength", heal_wait_qlen, options,
+                          uint32, failed);
+        ec_configure_background_heal_opts (ec, background_heals,
+                                           heal_wait_qlen);
         return 0;
 failed:
         return -1;
@@ -577,6 +597,10 @@ init (xlator_t *this)
     ec_method_initialize();
     GF_OPTION_INIT ("self-heal-daemon", ec->shd.enabled, bool, failed);
     GF_OPTION_INIT ("iam-self-heal-daemon", ec->shd.iamshd, bool, failed);
+    GF_OPTION_INIT ("background-heals", ec->background_heals, uint32, failed);
+    GF_OPTION_INIT ("heal-wait-qlength", ec->heal_wait_qlen, uint32, failed);
+    ec_configure_background_heal_opts (ec, ec->background_heals,
+                                       ec->heal_wait_qlen);
 
     if (ec->shd.iamshd)
             ec_selfheal_daemon_init (this);
@@ -1188,6 +1212,10 @@ int32_t ec_dump_private(xlator_t *this)
     gf_proc_dump_write("childs_up", "%u", ec->xl_up_count);
     gf_proc_dump_write("childs_up_mask", "%s",
                        ec_bin(tmp, sizeof(tmp), ec->xl_up, ec->nodes));
+    gf_proc_dump_write("background-heals", "%d", ec->background_heals);
+    gf_proc_dump_write("heal-wait-qlength", "%d", ec->heal_wait_qlen);
+    gf_proc_dump_write("healers", "%d", ec->healers);
+    gf_proc_dump_write("heal-waiters", "%d", ec->heal_waiters);
 
     return 0;
 }
@@ -1271,5 +1299,21 @@ struct volume_options options[] =
                      "translator is running as part of self-heal-daemon "
                      "or not."
     },
+    { .key = {"background-heals"},
+      .type = GF_OPTION_TYPE_INT,
+      .min = 0,/*Disabling background heals*/
+      .max = 256,
+      .default_value = "8",
+      .description = "This option can be used to control number of parallel"
+                     " heals",
+    },
+    { .key = {"heal-wait-qlength"},
+      .type = GF_OPTION_TYPE_INT,
+      .min = 0,
+      .max = 65536, /*Around 100MB as of now with sizeof(ec_fop_data_t) at 1800*/
+      .default_value = "128",
+      .description = "This option can be used to control number of heals"
+                     " that can wait",
+    },
     { }
 };
diff --git a/xlators/cluster/ec/src/ec.h b/xlators/cluster/ec/src/ec.h
index 7f140204ece..f335fd52afc 100644
--- a/xlators/cluster/ec/src/ec.h
+++ b/xlators/cluster/ec/src/ec.h
@@ -47,6 +47,8 @@ struct _ec
     gf_lock_t         lock;
     gf_timer_t *      timer;
     gf_boolean_t      shutdown;
+    uint32_t          background_heals;
+    uint32_t          heal_wait_qlen;
     struct list_head  pending_fops;
     struct list_head  heal_waiting;
     struct list_head  healing;