[sheepdog] [PATCH] sheep: add a kill node operation

[Liu Yuan]

From: Liu Yuan <tailai.ly at taobao.com>

This command is supposed to shut down the specified node correctly

usage:
 $ collie node kill node_id

Signed-off-by: Liu Yuan <tailai.ly at taobao.com>
---
 collie/node.c            |   37 +++++++++++++++++++++++++++++++++++++
 include/internal_proto.h |    1 +
 sheep/ops.c              |   14 ++++++++++++++
 sheep/request.c          |    1 +
 4 files changed, 53 insertions(+)

diff --git a/collie/node.c b/collie/node.c
index a778605..e8e6cac 100644
--- a/collie/node.c
+++ b/collie/node.c
@@ -157,7 +157,44 @@ static int node_recovery(int argc, char **argv)
 	return EXIT_SUCCESS;
 }
 
+static int node_kill(int argc, char **argv)
+{
+	char host[128];
+	int fd, node_id, ret;
+	unsigned wlen, rlen;
+	struct sd_node_req req;
+	struct sd_node_rsp *rsp = (struct sd_node_rsp *)&req;
+
+	node_id = strtol(argv[optind++], NULL, 10);
+	if (node_id < 0 || node_id >= sd_nodes_nr) {
+		fprintf(stderr, "Invalid node id '%d'\n", node_id);
+		exit(EXIT_USAGE);
+	}
+
+	addr_to_str(host, sizeof(host), sd_nodes[node_id].nid.addr, 0);
+
+	fd = connect_to(host, sd_nodes[node_id].nid.port);
+	if (fd < 0)
+		return EXIT_FAILURE;
+
+	sd_init_req((struct sd_req *)&req, SD_OP_KILL);
+
+	wlen = 0;
+	rlen = 0;
+	ret = exec_req(fd, (struct sd_req *)&req, NULL, &wlen, &rlen);
+	close(fd);
+
+	if (ret || rsp->result != SD_RES_SUCCESS) {
+		fprintf(stderr, "Failed to execute request\n");
+		exit(EXIT_FAILURE);
+	}
+
+	return EXIT_SUCCESS;
+}
+
 static struct subcommand node_cmd[] = {
+	{"kill", "<node id>", "aprh", "kill node",
+	 SUBCMD_FLAG_NEED_NODELIST|SUBCMD_FLAG_NEED_THIRD_ARG, node_kill},
 	{"list", NULL, "aprh", "list nodes",
 	 SUBCMD_FLAG_NEED_NODELIST, node_list},
 	{"info", NULL, "aprh", "show information about each node",
diff --git a/include/internal_proto.h b/include/internal_proto.h
index 584f41e..ff37b48 100644
--- a/include/internal_proto.h
+++ b/include/internal_proto.h
@@ -52,6 +52,7 @@
 #define SD_OP_TRACE_CAT      0x96
 #define SD_OP_STAT_RECOVERY  0x97
 #define SD_OP_FLUSH_DEL_CACHE  0x98
+#define SD_OP_KILL           0x99
 #define SD_OP_GET_OBJ_LIST   0xA1
 #define SD_OP_GET_EPOCH      0xA2
 #define SD_OP_CREATE_AND_WRITE_PEER 0xA3
diff --git a/sheep/ops.c b/sheep/ops.c
index 27dbdfa..b7346d1 100644
--- a/sheep/ops.c
+++ b/sheep/ops.c
@@ -548,6 +548,13 @@ static int local_trace_cat_ops(const struct sd_req *req, struct sd_rsp *rsp, voi
 	return SD_RES_SUCCESS;
 }
 
+static int local_kill(const struct sd_req *req, struct sd_rsp *rsp, void *data)
+{
+	sys_stat_set(SD_STATUS_SHUTDOWN);
+
+	return SD_RES_SUCCESS;
+}
+
 static int read_copy_from_replica(struct vnode_info *vnodes, uint32_t epoch,
 				  uint64_t oid, char *buf)
 {
@@ -921,6 +928,13 @@ static struct sd_op_template sd_ops[] = {
 		.process_main = local_trace_cat_ops,
 	},
 
+	[SD_OP_KILL] = {
+		.name = "KILL",
+		.type = SD_OP_TYPE_LOCAL,
+		.force = 1,
+		.process_main = local_kill,
+	},
+
 	/* gateway I/O operations */
 	[SD_OP_CREATE_AND_WRITE_OBJ] = {
 		.name = "CREATE_AND_WRITE_OBJ",
diff --git a/sheep/request.c b/sheep/request.c
index 35ac488..a61819b 100644
--- a/sheep/request.c
+++ b/sheep/request.c
@@ -77,6 +77,7 @@ static void gateway_op_done(struct work *work)
 	case SD_RES_NETWORK_ERROR:
 	case SD_RES_WAIT_FOR_JOIN:
 	case SD_RES_WAIT_FOR_FORMAT:
+	case SD_RES_SHUTDOWN:
 		dprintf("retrying failed I/O request "
 			"op %s result %d epoch %d, sys epoch %d\n",
 			op_name(req->op),
-- 
1.7.10.2