cluster/tier: do not block in synctask created from pause tier

We had run sleep() in the pause tier callback. Blocking within
a synctask is dangerous. The sleep() call does not inform
the synctask scheduler that a thread is no longer running.
It therefore believes it is running. If a second synctask already
exists, it may not be able to run. This occurs if the thread
limit in the pool has been reached.

Note the pool size only grows when a synctask is created, not
when it is moved from wait state to run state, as is the case
when an FOP completes. When the tier is paused during migration,
synctasks already exist waiting for responses to FOPs to the
server with high probability.

The fix is to yield() in the RPC callback, which will place
the synctask into the wait queue and free up a thread for the
FOP callback. A timer wakes the callback after sufficient
time has elapsed for the pause to occur.

This is a backport of 12987.

> Change-Id: I6a947ee04c6e5649946cb6d8207ba17263a67fc6
> BUG: 1267950
> Signed-off-by: Dan Lambright <dlambrig@redhat.com>
> Reviewed-on: http://review.gluster.org/12987
> Tested-by: Gluster Build System <jenkins@build.gluster.com>
> Reviewed-by: Rajesh Joseph <rjoseph@redhat.com>
Signed-off-by: Dan Lambright <dlambrig@redhat.com>

Change-Id: I9bb7564d57d59abb98e89c8d54c8b1ff4a5fc3f3
BUG: 1274100
Reviewed-on: http://review.gluster.org/13078
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Dan Lambright <dlambrig@redhat.com>
Tested-by: Dan Lambright <dlambrig@redhat.com>

3 files changed, 126 insertions, 18 deletions

diff --git a/xlators/cluster/dht/src/dht-common.h b/xlators/cluster/dht/src/dht-common.h
index 4e185c73bc5..6eec6c43b5d 100644
--- a/xlators/cluster/dht/src/dht-common.h
+++ b/xlators/cluster/dht/src/dht-common.h
@@ -21,6 +21,7 @@
 #include "libxlator.h"
 #include "syncop.h"
 #include "refcount.h"
+#include "timer.h"
 
 #ifndef _DHT_H
 #define _DHT_H
@@ -361,6 +362,9 @@ typedef struct gf_tier_conf {
         uint64_t                     st_last_demoted_size;
         int                          request_pause;
         gf_boolean_t                 paused;
+        struct synctask             *pause_synctask;
+        gf_timer_t                  *pause_timer;
+        pthread_mutex_t              pause_mutex;
 } gf_tier_conf_t;
 
 struct gf_defrag_info_ {
@@ -992,6 +996,9 @@ gf_defrag_status_get (gf_defrag_info_t *defrag, dict_t *dict);
 int
 gf_defrag_pause_tier (xlator_t *this, gf_defrag_info_t *defrag);
 
+void
+gf_defrag_wake_pause_tier (gf_tier_conf_t *defrag, gf_boolean_t pause);
+
 int
 gf_defrag_resume_tier (xlator_t *this, gf_defrag_info_t *defrag);
 
diff --git a/xlators/cluster/dht/src/dht-rebalance.c b/xlators/cluster/dht/src/dht-rebalance.c
index ce607011dae..a1860ccd900 100644
--- a/xlators/cluster/dht/src/dht-rebalance.c
+++ b/xlators/cluster/dht/src/dht-rebalance.c
@@ -3486,31 +3486,88 @@ out:
         return 0;
 }
 
+void
+gf_defrag_wake_pause_tier (gf_tier_conf_t *tier_conf, gf_boolean_t pause)
+{
+        int woke = 0;
+
+        pthread_mutex_lock (&tier_conf->pause_mutex);
+        if (tier_conf->pause_synctask) {
+                tier_conf->paused = pause;
+                synctask_wake (tier_conf->pause_synctask);
+                tier_conf->pause_synctask = 0;
+                woke = 1;
+        }
+        pthread_mutex_unlock (&tier_conf->pause_mutex);
+        tier_conf->request_pause = 0;
+
+        gf_msg ("tier", GF_LOG_DEBUG, 0,
+                DHT_MSG_TIER_PAUSED,
+                "woken %d paused %d", woke, tier_conf->paused);
+}
+
+void
+gf_defrag_pause_tier_timeout (void *data)
+{
+        xlator_t         *this                  = NULL;
+        dht_conf_t       *conf                  = NULL;
+        gf_defrag_info_t *defrag                = NULL;
+
+        this   = (xlator_t *) data;
+        GF_VALIDATE_OR_GOTO ("tier", this, out);
+
+        conf   = this->private;
+        GF_VALIDATE_OR_GOTO (this->name, conf, out);
+
+        defrag = conf->defrag;
+        GF_VALIDATE_OR_GOTO (this->name, defrag, out);
+
+        gf_msg (this->name, GF_LOG_DEBUG, 0,
+                DHT_MSG_TIER_PAUSED,
+                "Request pause timer timeout");
+
+        gf_defrag_wake_pause_tier (&defrag->tier_conf, _gf_false);
+
+out:
+        return;
+}
+
 int
 gf_defrag_pause_tier (xlator_t *this, gf_defrag_info_t *defrag)
 {
-        int          poll           = 0;
-        int          ret            = 0;
-        int          usec_sleep     = 100000;  /* 1/10th of a sec */
-        int          poll_max       = 15;      /* 15 times = wait at most 3/2 sec */
+        int             ret         = 0;
+        struct timespec delta       = {0,};
+        int             delay       = 2;
 
         if (defrag->defrag_status != GF_DEFRAG_STATUS_STARTED)
                 goto out;
 
         /*
-         * Set flag requesting to pause tiering. Wait a finite time for
+         * Set flag requesting to pause tiering. Wait 'delay' seconds for
          * tiering to actually stop as indicated by the "paused" boolean,
          * before returning success or failure.
          */
         defrag->tier_conf.request_pause = 1;
 
-        for (poll = 0; poll < poll_max; poll++) {
-                if ((defrag->tier_conf.paused == _gf_true) ||
-                    (defrag->defrag_status != GF_DEFRAG_STATUS_STARTED)) {
-                        goto out;
-                }
-                usleep (usec_sleep);
-        }
+        if (defrag->tier_conf.paused == _gf_true)
+                goto out;
+
+        gf_msg (this->name, GF_LOG_DEBUG, 0,
+                DHT_MSG_TIER_PAUSED,
+                "Request pause tier");
+
+        defrag->tier_conf.pause_synctask = synctask_get ();
+        delta.tv_sec  = delay;
+        delta.tv_nsec = 0;
+        defrag->tier_conf.pause_timer =
+                gf_timer_call_after (this->ctx, delta,
+                                     gf_defrag_pause_tier_timeout,
+                                     this);
+
+        synctask_yield (defrag->tier_conf.pause_synctask);
+
+        if (defrag->tier_conf.paused == _gf_true)
+                goto out;
 
         ret = -1;
 
@@ -3528,7 +3585,7 @@ gf_defrag_resume_tier (xlator_t *this, gf_defrag_info_t *defrag)
 {
         gf_msg (this->name, GF_LOG_DEBUG, 0,
                 DHT_MSG_TIER_RESUME,
-                "Resume tiering");
+                "Pause end. Resume tiering");
 
         defrag->tier_conf.request_pause = 0;
         defrag->tier_conf.paused = _gf_false;
diff --git a/xlators/cluster/dht/src/tier.c b/xlators/cluster/dht/src/tier.c
index 393dde4361c..944d3af50c3 100644
--- a/xlators/cluster/dht/src/tier.c
+++ b/xlators/cluster/dht/src/tier.c
@@ -1375,10 +1375,8 @@ tier_start (xlator_t *this, gf_defrag_info_t *defrag)
                         goto out;
                 }
 
-                if (defrag->tier_conf.request_pause)
-                        defrag->tier_conf.paused = _gf_true;
-                else
-                        defrag->tier_conf.paused = _gf_false;
+                if (tier_conf->request_pause)
+                        gf_defrag_wake_pause_tier (tier_conf, _gf_true);
 
                 sleep(1);
 
@@ -1809,6 +1807,8 @@ tier_init (xlator_t *this)
 
         defrag->tier_conf.request_pause = 0;
 
+        pthread_mutex_init (&defrag->tier_conf.pause_mutex, 0);
+
         ret = dict_get_str (this->options,
                               "tier-pause", &paused);
 
@@ -1877,6 +1877,40 @@ out:
 
 
 int
+tier_cli_pause_done (int op_ret, call_frame_t *sync_frame, void *data)
+{
+        gf_msg ("tier", GF_LOG_INFO, 0,
+                DHT_MSG_TIER_PAUSED,
+                "Migrate file paused with op_ret %d", op_ret);
+
+        return op_ret;
+}
+
+int
+tier_cli_pause (void *data)
+{
+        gf_defrag_info_t        *defrag         = NULL;
+        xlator_t                *this           = NULL;
+        dht_conf_t              *conf           = NULL;
+        int                      ret            = -1;
+
+        this = data;
+
+        conf = this->private;
+        GF_VALIDATE_OR_GOTO (this->name, conf, exit);
+
+        defrag = conf->defrag;
+        GF_VALIDATE_OR_GOTO (this->name, defrag, exit);
+
+        gf_defrag_pause_tier (this, defrag);
+
+        ret = 0;
+exit:
+        return ret;
+}
+
+
+int
 tier_reconfigure (xlator_t *this, dict_t *options)
 {
         dht_conf_t       *conf           = NULL;
@@ -1885,6 +1919,7 @@ tier_reconfigure (xlator_t *this, dict_t *options)
         int               migrate_mb     = 0;
         gf_boolean_t      req_pause      = _gf_false;
         int               ret            = 0;
+        call_frame_t            *frame  = NULL;
 
         conf = this->private;
 
@@ -1933,7 +1968,16 @@ tier_reconfigure (xlator_t *this, dict_t *options)
                                   bool, out);
 
                 if (req_pause == _gf_true) {
-                        ret = gf_defrag_pause_tier (this, defrag);
+
+                        frame = create_frame (this, this->ctx->pool);
+                        if (!frame)
+                                goto out;
+
+                        frame->root->pid = GF_CLIENT_PID_DEFRAG;
+
+                        ret = synctask_new (this->ctx->env, tier_cli_pause,
+                                            tier_cli_pause_done, frame, this);
+
                         if (ret) {
                                 gf_msg (this->name, GF_LOG_ERROR, 0,
                                         DHT_MSG_LOG_TIER_ERROR,