ceph: reset osd after relevant messages timed out

This simplifies the process of timing out messages. We
keep lru of current messages that are in flight. If a
timeout has passed, we reset the osd connection, so that
messages will be retransmitted.  This is a failsafe in case
we hit some sort of problem sending out message to the OSD.
Normally, we'll get notification via an updated osdmap if
there are problems.

If a request is older than the keepalive timeout, send a
keepalive to ensure we detect any breaks in the TCP connection.
Signed-off-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>

fs/ceph/osd_client.c

@@ -17,6 +17,8 @@
 #define OSD_OPREPLY_FRONT_LEN	512
 
 const static struct ceph_connection_operations osd_con_ops;
+static int __kick_requests(struct ceph_osd_client *osdc,
+			  struct ceph_osd *kickosd);
 
 static void kick_requests(struct ceph_osd_client *osdc, struct ceph_osd *osd);
 
@@ -339,6 +341,7 @@ static struct ceph_osd *create_osd(struct ceph_osd_client *osdc)
 	osd->o_con.ops = &osd_con_ops;
 	osd->o_con.peer_name.type = CEPH_ENTITY_TYPE_OSD;
 
+	INIT_LIST_HEAD(&osd->o_keepalive_item);
 	return osd;
 }
 
@@ -461,6 +464,16 @@ static struct ceph_osd *__lookup_osd(struct ceph_osd_client *osdc, int o)
 	return NULL;
 }
 
+static void __schedule_osd_timeout(struct ceph_osd_client *osdc)
+{
+	schedule_delayed_work(&osdc->timeout_work,
+			osdc->client->mount_args->osd_keepalive_timeout * HZ);
+}
+
+static void __cancel_osd_timeout(struct ceph_osd_client *osdc)
+{
+	cancel_delayed_work(&osdc->timeout_work);
+}
 
 /*
  * Register request, assign tid.  If this is the first request, set up
@@ -472,21 +485,16 @@ static void register_request(struct ceph_osd_client *osdc,
 	mutex_lock(&osdc->request_mutex);
 	req->r_tid = ++osdc->last_tid;
 	req->r_request->hdr.tid = cpu_to_le64(req->r_tid);
+	INIT_LIST_HEAD(&req->r_req_lru_item);
 
 	dout("register_request %p tid %lld\n", req, req->r_tid);
 	__insert_request(osdc, req);
 	ceph_osdc_get_request(req);
 	osdc->num_requests++;
 
-	req->r_timeout_stamp =
-		jiffies + osdc->client->mount_args->osd_timeout*HZ;
-
 	if (osdc->num_requests == 1) {
-		osdc->timeout_tid = req->r_tid;
-		dout("  timeout on tid %llu at %lu\n", req->r_tid,
-		     req->r_timeout_stamp);
-		schedule_delayed_work(&osdc->timeout_work,
-		      round_jiffies_relative(req->r_timeout_stamp - jiffies));
+		dout(" first request, scheduling timeout\n");
+		__schedule_osd_timeout(osdc);
 	}
 	mutex_unlock(&osdc->request_mutex);
 }
@@ -513,21 +521,10 @@ static void __unregister_request(struct ceph_osd_client *osdc,
 
 	ceph_osdc_put_request(req);
 
-	if (req->r_tid == osdc->timeout_tid) {
-		if (osdc->num_requests == 0) {
-			dout("no requests, canceling timeout\n");
-			osdc->timeout_tid = 0;
-			cancel_delayed_work(&osdc->timeout_work);
-		} else {
-			req = rb_entry(rb_first(&osdc->requests),
-				       struct ceph_osd_request, r_node);
-			osdc->timeout_tid = req->r_tid;
-			dout("rescheduled timeout on tid %llu at %lu\n",
-			     req->r_tid, req->r_timeout_stamp);
-			schedule_delayed_work(&osdc->timeout_work,
-			      round_jiffies_relative(req->r_timeout_stamp -
-						     jiffies));
-		}
+	list_del_init(&req->r_req_lru_item);
+	if (osdc->num_requests == 0) {
+		dout(" no requests, canceling timeout\n");
+		__cancel_osd_timeout(osdc);
 	}
 }
 
@@ -540,6 +537,7 @@ static void __cancel_request(struct ceph_osd_request *req)
 		ceph_con_revoke(&req->r_osd->o_con, req->r_request);
 		req->r_sent = 0;
 	}
+	list_del_init(&req->r_req_lru_item);
 }
 
 /*
@@ -635,7 +633,8 @@ static int __send_request(struct ceph_osd_client *osdc,
 	reqhead->flags |= cpu_to_le32(req->r_flags);  /* e.g., RETRY */
 	reqhead->reassert_version = req->r_reassert_version;
 
-	req->r_timeout_stamp = jiffies+osdc->client->mount_args->osd_timeout*HZ;
+	req->r_sent_stamp = jiffies;
+	list_move_tail(&osdc->req_lru, &req->r_req_lru_item);
 
 	ceph_msg_get(req->r_request); /* send consumes a ref */
 	ceph_con_send(&req->r_osd->o_con, req->r_request);
@@ -656,11 +655,14 @@ static void handle_timeout(struct work_struct *work)
 {
 	struct ceph_osd_client *osdc =
 		container_of(work, struct ceph_osd_client, timeout_work.work);
-	struct ceph_osd_request *req;
+	struct ceph_osd_request *req, *last_req = NULL;
 	struct ceph_osd *osd;
 	unsigned long timeout = osdc->client->mount_args->osd_timeout * HZ;
-	unsigned long next_timeout = timeout + jiffies;
+	unsigned long keepalive =
+		osdc->client->mount_args->osd_keepalive_timeout * HZ;
+	unsigned long last_sent = 0;
 	struct rb_node *p;
+	struct list_head slow_osds;
 
 	dout("timeout\n");
 	down_read(&osdc->map_sem);
@@ -683,25 +685,56 @@ static void handle_timeout(struct work_struct *work)
 			continue;
 		}
 	}
-	for (p = rb_first(&osdc->osds); p; p = rb_next(p)) {
-		osd = rb_entry(p, struct ceph_osd, o_node);
-		if (list_empty(&osd->o_requests))
-			continue;
-		req = list_first_entry(&osd->o_requests,
-				       struct ceph_osd_request, r_osd_item);
-		if (time_before(jiffies, req->r_timeout_stamp))
-			continue;
 
-		dout(" tid %llu (at least) timed out on osd%d\n",
+	/*
+	 * reset osds that appear to be _really_ unresponsive.  this
+	 * is a failsafe measure.. we really shouldn't be getting to
+	 * this point if the system is working properly.  the monitors
+	 * should mark the osd as failed and we should find out about
+	 * it from an updated osd map.
+	 */
+	while (!list_empty(&osdc->req_lru)) {
+		req = list_entry(osdc->req_lru.next, struct ceph_osd_request,
+				 r_req_lru_item);
+
+		if (time_before(jiffies, req->r_sent_stamp + timeout))
+			break;
+
+		BUG_ON(req == last_req && req->r_sent_stamp == last_sent);
+		last_req = req;
+		last_sent = req->r_sent_stamp;
+
+		osd = req->r_osd;
+		BUG_ON(!osd);
+		pr_warning(" tid %llu timed out on osd%d, will reset osd\n",
+			   req->r_tid, osd->o_osd);
+		__kick_requests(osdc, osd);
+	}
+
+	/*
+	 * ping osds that are a bit slow.  this ensures that if there
+	 * is a break in the TCP connection we will notice, and reopen
+	 * a connection with that osd (from the fault callback).
+	 */
+	INIT_LIST_HEAD(&slow_osds);
+	list_for_each_entry(req, &osdc->req_lru, r_req_lru_item) {
+		if (time_before(jiffies, req->r_sent_stamp + keepalive))
+			break;
+
+		osd = req->r_osd;
+		BUG_ON(!osd);
+		dout(" tid %llu is slow, will send keepalive on osd%d\n",
 		     req->r_tid, osd->o_osd);
-		req->r_timeout_stamp = next_timeout;
+		list_move_tail(&osd->o_keepalive_item, &slow_osds);
+	}
+	while (!list_empty(&slow_osds)) {
+		osd = list_entry(slow_osds.next, struct ceph_osd,
+				 o_keepalive_item);
+		list_del_init(&osd->o_keepalive_item);
 		ceph_con_keepalive(&osd->o_con);
 	}
 
-	if (osdc->timeout_tid)
-		schedule_delayed_work(&osdc->timeout_work,
-				      round_jiffies_relative(timeout));
-
+	__schedule_osd_timeout(osdc);
 	mutex_unlock(&osdc->request_mutex);
 
 	up_read(&osdc->map_sem);
@@ -819,18 +852,7 @@ static void handle_reply(struct ceph_osd_client *osdc, struct ceph_msg *msg,
 }
 
 
-/*
- * Resubmit osd requests whose osd or osd address has changed.  Request
- * a new osd map if osds are down, or we are otherwise unable to determine
- * how to direct a request.
- *
- * Close connections to down osds.
- *
- * If @who is specified, resubmit requests for that specific osd.
- *
- * Caller should hold map_sem for read and request_mutex.
- */
-static void kick_requests(struct ceph_osd_client *osdc,
+static int __kick_requests(struct ceph_osd_client *osdc,
 			  struct ceph_osd *kickosd)
 {
 	struct ceph_osd_request *req;
@@ -839,7 +861,6 @@ static void kick_requests(struct ceph_osd_client *osdc,
 	int err;
 
 	dout("kick_requests osd%d\n", kickosd ? kickosd->o_osd : -1);
-	mutex_lock(&osdc->request_mutex);
 	if (kickosd) {
 		__reset_osd(osdc, kickosd);
 	} else {
@@ -900,14 +921,36 @@ static void kick_requests(struct ceph_osd_client *osdc,
 			req->r_resend = true;
 		}
 	}
+
+	return needmap;
+}
+
+/*
+ * Resubmit osd requests whose osd or osd address has changed.  Request
+ * a new osd map if osds are down, or we are otherwise unable to determine
+ * how to direct a request.
+ *
+ * Close connections to down osds.
+ *
+ * If @who is specified, resubmit requests for that specific osd.
+ *
+ * Caller should hold map_sem for read and request_mutex.
+ */
+static void kick_requests(struct ceph_osd_client *osdc,
+			  struct ceph_osd *kickosd)
+{
+	int needmap;
+
+	mutex_lock(&osdc->request_mutex);
+	needmap = __kick_requests(osdc, kickosd);
 	mutex_unlock(&osdc->request_mutex);
 
 	if (needmap) {
 		dout("%d requests for down osds, need new map\n", needmap);
 		ceph_monc_request_next_osdmap(&osdc->client->monc);
 	}
-}
 
+}
 /*
  * Process updated osd map.
  *
@@ -1164,11 +1207,11 @@ int ceph_osdc_init(struct ceph_osd_client *osdc, struct ceph_client *client)
 	init_completion(&osdc->map_waiters);
 	osdc->last_requested_map = 0;
 	mutex_init(&osdc->request_mutex);
-	osdc->timeout_tid = 0;
 	osdc->last_tid = 0;
 	osdc->osds = RB_ROOT;
 	INIT_LIST_HEAD(&osdc->osd_lru);
 	osdc->requests = RB_ROOT;
+	INIT_LIST_HEAD(&osdc->req_lru);
 	osdc->num_requests = 0;
 	INIT_DELAYED_WORK(&osdc->timeout_work, handle_timeout);
 	INIT_DELAYED_WORK(&osdc->osds_timeout_work, handle_osds_timeout);

fs/ceph/osd_client.h

@@ -36,12 +36,15 @@ struct ceph_osd {
 	void *o_authorizer_buf, *o_authorizer_reply_buf;
 	size_t o_authorizer_buf_len, o_authorizer_reply_buf_len;
 	unsigned long lru_ttl;
+	int o_marked_for_keepalive;
+	struct list_head o_keepalive_item;
 };
 
 /* an in-flight request */
 struct ceph_osd_request {
 	u64             r_tid;              /* unique for this client */
 	struct rb_node  r_node;
+	struct list_head r_req_lru_item;
 	struct list_head r_osd_item;
 	struct ceph_osd *r_osd;
 	struct ceph_pg   r_pgid;
@@ -67,7 +70,7 @@ struct ceph_osd_request {
 
 	char              r_oid[40];          /* object name */
 	int               r_oid_len;
-	unsigned long     r_timeout_stamp;
+	unsigned long     r_sent_stamp;
 	bool              r_resend;           /* msg send failed, needs retry */
 
 	struct ceph_file_layout r_file_layout;
@@ -92,6 +95,7 @@ struct ceph_osd_client {
 	u64                    timeout_tid;   /* tid of timeout triggering rq */
 	u64                    last_tid;      /* tid of last request */
 	struct rb_root         requests;      /* pending requests */
+	struct list_head       req_lru;	      /* pending requests lru */
 	int                    num_requests;
 	struct delayed_work    timeout_work;
 	struct delayed_work    osds_timeout_work;

fs/ceph/super.c

@@ -292,6 +292,7 @@ enum {
 	Opt_wsize,
 	Opt_rsize,
 	Opt_osdtimeout,
+	Opt_osdkeepalivetimeout,
 	Opt_mount_timeout,
 	Opt_osd_idle_ttl,
 	Opt_caps_wanted_delay_min,
@@ -322,6 +323,7 @@ static match_table_t arg_tokens = {
 	{Opt_wsize, "wsize=%d"},
 	{Opt_rsize, "rsize=%d"},
 	{Opt_osdtimeout, "osdtimeout=%d"},
+	{Opt_osdkeepalivetimeout, "osdkeepalive=%d"},
 	{Opt_mount_timeout, "mount_timeout=%d"},
 	{Opt_osd_idle_ttl, "osd_idle_ttl=%d"},
 	{Opt_caps_wanted_delay_min, "caps_wanted_delay_min=%d"},
@@ -367,7 +369,8 @@ static struct ceph_mount_args *parse_mount_args(int flags, char *options,
 	/* start with defaults */
 	args->sb_flags = flags;
 	args->flags = CEPH_OPT_DEFAULT;
-	args->osd_timeout = 5;    /* seconds */
+	args->osd_timeout = CEPH_OSD_TIMEOUT_DEFAULT;
+	args->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT;
 	args->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT; /* seconds */
 	args->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT;   /* seconds */
 	args->caps_wanted_delay_min = CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT;
@@ -468,6 +471,9 @@ static struct ceph_mount_args *parse_mount_args(int flags, char *options,
 		case Opt_osdtimeout:
 			args->osd_timeout = intval;
 			break;
+		case Opt_osdkeepalivetimeout:
+			args->osd_keepalive_timeout = intval;
+			break;
 		case Opt_mount_timeout:
 			args->mount_timeout = intval;
 			break;

fs/ceph/super.h

@@ -62,6 +62,7 @@ struct ceph_mount_args {
 	int max_readdir;      /* max readdir size */
 	int congestion_kb;      /* max readdir size */
 	int osd_timeout;
+	int osd_keepalive_timeout;
 	char *snapdir_name;   /* default ".snap" */
 	char *name;
 	char *secret;
@@ -72,6 +73,8 @@ struct ceph_mount_args {
  * defaults
  */
 #define CEPH_MOUNT_TIMEOUT_DEFAULT  60
+#define CEPH_OSD_TIMEOUT_DEFAULT    60  /* seconds */
+#define CEPH_OSD_KEEPALIVE_DEFAULT  5
 #define CEPH_OSD_IDLE_TTL_DEFAULT    60
 #define CEPH_MOUNT_RSIZE_DEFAULT    (512*1024) /* readahead */