[sheepdog] [PATCH v3 11/12] collie: implement "collie cluster snapshot" subcommand

[Kai Zhang]

Command example:
collie cluster snapshot save /localpath

collie cluster snapshot list /localpath

collie cluster snapshot load /localpath

Signed-off-by: Kai Zhang <kyle at zelin.io>
---
 collie/cluster.c |  132 +++++++++++++++++++++++++++++++++++++++++++++++++++---
 1 files changed, 126 insertions(+), 6 deletions(-)

diff --git a/collie/cluster.c b/collie/cluster.c
index fa2e131..9a0cf6a 100644
--- a/collie/cluster.c
+++ b/collie/cluster.c
@@ -15,14 +15,15 @@
 #include <sys/time.h>
 
 #include "collie.h"
+#include "farm/farm.h"
 
 static struct sd_option cluster_options[] = {
 	{'b', "store", true, "specify backend store"},
 	{'c', "copies", true, "specify the default data redundancy (number of copies)"},
 	{'m', "mode", true, "mode (safe, quorum, unsafe)"},
 	{'f', "force", false, "do not prompt for confirmation"},
-	{'R', "restore", true, "restore the cluster"},
-	{'l', "list", false, "list the user epoch information"},
+	{'e', "epoch", true, "specify epoch to be loaded to new cluster."
+			     " default is to load the last epoch"},
 
 	{ 0, NULL, false, NULL },
 };
@@ -223,6 +224,110 @@ static int cluster_shutdown(int argc, char **argv)
 	return EXIT_SUCCESS;
 }
 
+static void print_list(void *buf, unsigned len)
+{
+	struct snap_log *log_buf = (struct snap_log *)buf;
+	unsigned nr = len / sizeof(struct snap_log), i;
+
+	printf("Index\t\tSnapshot Time\n");
+	for (i = 0; i < nr; i++, log_buf++) {
+		time_t *t = (time_t *)&log_buf->time;
+		printf("%d\t\t", log_buf->epoch);
+		printf("%s", ctime(t));
+	}
+}
+
+static int list_snapshot(int argc, char **argv)
+{
+	char *path = argv[optind++];
+	void *buf = NULL;
+	int log_nr;
+	int ret = EXIT_SYSFAIL;
+
+	if (farm_init(path) != SD_RES_SUCCESS)
+		goto out;
+
+	buf = snap_log_read(&log_nr);
+	if (!buf)
+		goto out;
+
+	print_list(buf, log_nr * sizeof(struct snap_log));
+	ret = EXIT_SUCCESS;
+out:
+	if (ret)
+		sd_eprintf("fail to list snapshot");
+	free(buf);
+	return ret;
+}
+
+static void fill_obj_tree(uint32_t vid, const char *name, const char *tag,
+			  uint32_t snapid, uint32_t flags,
+			  const struct sd_inode *i, void *data)
+{
+	uint64_t vdi_oid = vid_to_vdi_oid(vid);
+
+	/* ignore active vdi */
+	if (!i->snap_ctime)
+		return;
+
+	obj_tree_insert(vdi_oid, VDI, i->nr_copies);
+
+	for (int idx = 0; idx < MAX_DATA_OBJS; idx++) {
+		if (i->data_vdi_id[idx]) {
+			uint64_t oid = vid_to_data_oid(i->data_vdi_id[idx],
+						       idx);
+			obj_tree_insert(oid, OBJECT, i->nr_copies);
+		}
+	}
+}
+
+static int save_snapshot(int argc, char **argv)
+{
+	char *path = argv[optind++];
+	int ret = EXIT_SYSFAIL;
+
+	sd_dprintf("saving cluster snapshop to path: %s", path);
+
+	if (parse_vdi(fill_obj_tree, SD_INODE_SIZE, NULL) != SD_RES_SUCCESS)
+		goto out;
+
+	if (farm_init(path) != SD_RES_SUCCESS)
+		goto out;
+
+	if (farm_save_snapshot() != SD_RES_SUCCESS)
+		goto out;
+
+	ret = EXIT_SUCCESS;
+out:
+	if (ret)
+		sd_eprintf("fail to save snapshot to path: %s", path);
+	free_obj_tree();
+	return ret;
+}
+
+static int load_snapshot(int argc, char **argv)
+{
+	char *path = argv[optind++];
+	int ret = EXIT_SYSFAIL;
+
+	sd_dprintf("loading cluster snapshot from path: %s", path);
+
+	if (farm_init(path) != SD_RES_SUCCESS)
+		goto out;
+
+	if (cluster_format(0, NULL) != SD_RES_SUCCESS)
+		goto out;
+
+	if (farm_load_snapshot(cluster_cmd_data.epoch) != SD_RES_SUCCESS)
+		goto out;
+
+	ret = EXIT_SUCCESS;
+out:
+	if (ret)
+		sd_eprintf("fail to load snapshot");
+	return ret;
+}
+
 #define RECOVER_PRINT \
 	"Caution! Please try starting all the cluster nodes normally before\n" \
 	"running this command.\n\n" \
@@ -315,6 +420,21 @@ static int cluster_recover(int argc, char **argv)
 	return do_generic_subcommand(cluster_recover_cmd, argc, argv);
 }
 
+/* Subcommand list of snapshot */
+static struct subcommand cluster_snapshot_cmd[] = {
+	{"save", NULL, "eh", "save snapshot to localpath",
+	 NULL, 0, save_snapshot, cluster_options},
+	{"list", NULL, "eh", "list snapshot of localpath",
+	 NULL, 0, list_snapshot, NULL},
+	{"load", NULL, "eh", "list snapshot of localpath",
+	 NULL, 0, load_snapshot, NULL}
+};
+
+static int cluster_snapshot(int argc, char **argv)
+{
+	return do_generic_subcommand(cluster_snapshot_cmd, argc, argv);
+}
+
 static struct subcommand cluster_cmd[] = {
 	{"info", NULL, "aprh", "show cluster information",
 	 NULL, SUBCMD_FLAG_NEED_NODELIST, cluster_info, cluster_options},
@@ -322,6 +442,9 @@ static struct subcommand cluster_cmd[] = {
 	 NULL, 0, cluster_format, cluster_options},
 	{"shutdown", NULL, "aph", "stop Sheepdog",
 	 NULL, 0, cluster_shutdown, cluster_options},
+	{"snapshot", "<path>", "aeph", "snapshot/restore the cluster",
+	 cluster_snapshot_cmd, SUBCMD_FLAG_NEED_ARG,
+	 cluster_snapshot, cluster_options},
 	{"recover", NULL, "afph",
 	 "See 'collie cluster recover' for more information\n",
 	 cluster_recover_cmd, SUBCMD_FLAG_NEED_ARG,
@@ -369,7 +492,7 @@ static int cluster_parser(int ch, char *opt)
 	case 'f':
 		cluster_cmd_data.force = true;
 		break;
-	case 'R':
+	case 'e':
 		cluster_cmd_data.epoch = strtol(opt, &p, 10);
 		if (opt == p) {
 			fprintf(stderr, "The epoch must be an integer\n");
@@ -380,9 +503,6 @@ static int cluster_parser(int ch, char *opt)
 			exit(EXIT_FAILURE);
 		}
 		break;
-	case 'l':
-		cluster_cmd_data.list = true;
-		break;
 	}
 
 	return 0;
-- 
1.7.1