fix(stale_snaps): cherry-pick to delete stale snaps and installs debu…

…g tools (#229)

* fix(stale_snaps): delete internal created snaps for helping in rebuil… (#227)

...d during data connection

Signed-off-by: Vitta <vitta@mayadata.io>

* [2502] feat(debug-tools): installing curl/tcpdump/nslookup/ping packages

Signed-off-by: Vitta <vitta@mayadata.io>

docker/Dockerfile.base

@@ -8,9 +8,11 @@ FROM openebs/cstor-ubuntu:xenial-20181005
 RUN apt-get clean && \
     rm -rf /var/lib/apt/lists/* && \ 
     apt-get update && apt-get install -y \
+    curl tcpdump dnsutils iputils-ping \
     libaio1 libaio-dev \
     libkqueue-dev libssl1.0.0 rsyslog net-tools gdb apt-utils \
     sed libjemalloc-dev openssh-server
+RUN apt-get -y install apt-file && apt-file update
 
 COPY zfs/bin/* /usr/local/bin/
 COPY zfs/lib/* /usr/lib/

include/data_conn.h

@@ -79,8 +79,6 @@ int uzfs_zvol_create_internal_snapshot(zvol_state_t *zv, zvol_state_t **snap_zv,
 void signal_fds_related_to_zinfo(zvol_info_t *zinfo);
 void quiesce_wait(zvol_info_t *zinfo);
 
-int uzfs_zvol_create_internal_snapshot(zvol_state_t *zv, zvol_state_t **snap_zv,
-    uint64_t io_num);
 int uzfs_zvol_handle_rebuild_snap_done(zvol_io_hdr_t *, int, zvol_info_t *);
 
 #ifdef __cplusplus

include/uzfs_rebuilding.h

@@ -66,6 +66,8 @@ int uzfs_zvol_release_internal_clone(zvol_state_t *zv,
  */
 int uzfs_destroy_all_internal_snapshots(zvol_state_t *zv);
 boolean_t is_stale_clone(zvol_state_t *);
+int uzfs_destroy_all_iosnap_snapshots(zvol_state_t *zv);
+boolean_t is_stale_clone(zvol_state_t *);
 
 #ifdef __cplusplus
 }

lib/libzpool/uzfs_rebuilding.c

@@ -394,12 +394,36 @@ uzfs_zvol_get_or_create_internal_clone(zvol_state_t *zv,
 	return (ret);
 }
 
+typedef boolean_t (*uzfs_snapname_matching_func)(char *snapname);
+
 /*
- * To destroy all internal created snapshot
- * on a dataset
+ * Return true for snaps that are internal created
  */
-int
-uzfs_destroy_all_internal_snapshots(zvol_state_t *zv)
+boolean_t
+uzfs_match_internal_snapshots(char *snapname)
+{
+	if ((strcmp(snapname, REBUILD_SNAPSHOT_SNAPNAME) == 0) ||
+	    (strncmp(snapname, IO_DIFF_SNAPNAME,
+	    sizeof (IO_DIFF_SNAPNAME) - 1) == 0))
+		return (B_TRUE);
+	return (B_FALSE);
+}
+
+/*
+ * Return true for snaps that starts with .io_snap
+ */
+boolean_t
+uzfs_match_iosnap_snapshots(char *snapname)
+{
+	if (strncmp(snapname, IO_DIFF_SNAPNAME,
+	    sizeof (IO_DIFF_SNAPNAME) - 1) == 0)
+		return (B_TRUE);
+	return (B_FALSE);
+}
+
+static int
+uzfs_destroy_matching_snapshots(zvol_state_t *zv,
+    uzfs_snapname_matching_func matching_fn)
 {
 	int ret;
 	char snapname[MAXNAMELEN];
@@ -423,11 +447,9 @@ uzfs_destroy_all_internal_snapshots(zvol_state_t *zv)
 			break;
 		}
 
-		if (!(strcmp(snapname, REBUILD_SNAPSHOT_SNAPNAME) == 0) &&
-		    !(strncmp(snapname, IO_DIFF_SNAPNAME,
-		    sizeof (IO_DIFF_SNAPNAME) - 1) == 0)) {
+		/* skip destroying non-matching snaps */
+		if ((*matching_fn)(snapname) == B_FALSE)
 			continue;
-		}
 
 		ret = destroy_snapshot_zv(zv, snapname);
 		if (ret != 0) {
@@ -439,3 +461,25 @@ uzfs_destroy_all_internal_snapshots(zvol_state_t *zv)
 
 	return (ret);
 }
+
+/*
+ * To destroy all internal created snapshot
+ * on a dataset
+ */
+int
+uzfs_destroy_all_internal_snapshots(zvol_state_t *zv)
+{
+	int ret;
+	ret = uzfs_destroy_matching_snapshots(zv,
+	    uzfs_match_internal_snapshots);
+	return (ret);
+}
+
+int
+uzfs_destroy_all_iosnap_snapshots(zvol_state_t *zv)
+{
+	int ret;
+	ret = uzfs_destroy_matching_snapshots(zv,
+	    uzfs_match_iosnap_snapshots);
+	return (ret);
+}

lib/libzrepl/data_conn.c

@@ -1520,6 +1520,7 @@ uzfs_zvol_rebuild_scanner(void *arg)
 	uint64_t	payload_size = 0;
 	char		*snap_name;
 	zvol_rebuild_scanner_info_t	*warg = NULL;
+	char		*at_ptr = NULL;
 
 	if ((rc = setsockopt(fd, SOL_SOCKET, SO_LINGER, &lo, sizeof (lo)))
 	    != 0) {
@@ -1706,6 +1707,11 @@ uzfs_zvol_rebuild_scanner(void *arg)
 					    snap_zv->zv_name);
 					LOG_INFO("closing snap %s", snap_name);
 					uzfs_close_dataset(snap_zv);
+					at_ptr = strchr(snap_name, '@');
+					VERIFY3P(at_ptr, !=, NULL);
+					VERIFY0(strncmp(at_ptr + 1,
+					    IO_DIFF_SNAPNAME,
+					    sizeof (IO_DIFF_SNAPNAME) - 1));
 #if DEBUG
 					if (inject_error.delay.
 					    helping_replica_rebuild_complete
@@ -1741,6 +1747,10 @@ uzfs_zvol_rebuild_scanner(void *arg)
 			snap_name = kmem_asprintf("%s", snap_zv->zv_name);
 			LOG_INFO("closing snap on conn break %s", snap_name);
 			uzfs_close_dataset(snap_zv);
+			at_ptr = strchr(snap_name, '@');
+			VERIFY3P(at_ptr, !=, NULL);
+			VERIFY0(strncmp(at_ptr + 1, IO_DIFF_SNAPNAME,
+			    sizeof (IO_DIFF_SNAPNAME) - 1));
 #if DEBUG
 			if (inject_error.delay.helping_replica_rebuild_complete
 			    == 1)
@@ -2197,6 +2207,9 @@ open_zvol(int fd, zvol_info_t **zinfopp)
 		goto error_ret;
 	}
 
+	/* Destroy snaps that are internally created to help during rebuild */
+	uzfs_destroy_all_iosnap_snapshots(zinfo->main_zv);
+
 	status = find_apt_zvol_status(zinfo, &open_data);
 	if (status == ZVOL_STATUS_HEALTHY) {
 		ASSERT3P(zinfo->snapshot_zv, ==, NULL);
@@ -2222,6 +2235,12 @@ open_zvol(int fd, zvol_info_t **zinfopp)
 			(void) pthread_mutex_unlock(&zinfo->zinfo_mutex);
 			goto open_reply;
 		}
+
+		/*
+		 * Destroy snaps that are internally created to help
+		 * during rebuild
+		 */
+		uzfs_destroy_all_iosnap_snapshots(zinfo->clone_zv);
 	}
 	/*
 	 * TODO: Once we support multiple concurrent data connections for a

tests/cbtest/gtest/test_zrepl_prot.cc

@@ -417,6 +417,58 @@ class Target {
 	int m_listenfd;
 };
 
+/*
+ * Stale Snapshot deletion verification
+ */
+TEST(StaleSnapshot, Destroy) {
+	Zrepl zrepl;
+	Target target;
+	int rc, control_fd;
+	SocketFd datasock;
+	TestPool pool("stale_snap_pool");
+	std::string vol_name = pool.getZvolName("vol");
+	std::string snap_name1 = pool.getZvolName("vol@usersnap");
+	std::string snap_name2 = pool.getZvolName("vol@.io_snap");
+	std::string snap_name3 = pool.getZvolName("vol@.io_snap1.2");
+	std::string snap_name4 = pool.getZvolName("vol@rebuild_snap");
+	std::string snap_name5 = pool.getZvolName("vol_rebuild_clone@.io_snap");
+	std::string snap_name6 = pool.getZvolName("vol_rebuild_clone@.io_snap1.2");
+	std::string host;
+	uint16_t port;
+	std::string output;
+
+	zrepl.start();
+	pool.create();
+	pool.createZvol("vol", "-o io.openebs:targetip=127.0.0.1");
+	pool.createZvol("vol_rebuild_clone", "-o io.openebs:targetip=127.0.0.2");
+	output = execCmd("zfs", std::string("snapshot ") + snap_name1);
+	output = execCmd("zfs", std::string("snapshot ") + snap_name2);
+	output = execCmd("zfs", std::string("snapshot ") + snap_name3);
+	output = execCmd("zfs", std::string("snapshot ") + snap_name4);
+	output = execCmd("zfs", std::string("snapshot ") + snap_name5);
+	output = execCmd("zfs", std::string("snapshot ") + snap_name6);
+	output = execCmd("zfs", std::string("list -t all"));
+
+	printf("%s\n", output.c_str());
+
+	zrepl.kill();
+
+	zrepl.start();
+	pool.import();
+
+	rc = target.listen();
+	ASSERT_GE(rc, 0);
+	control_fd = target.accept(-1);
+	ASSERT_GE(control_fd, 0);
+
+	do_handshake(vol_name, host, port, NULL, NULL, control_fd, ZVOL_OP_STATUS_OK);
+	do_data_connection(datasock.fd(), host, port, vol_name, 4096, 2);
+
+	output = execCmd("zfs", std::string("list -t all"));
+	printf("%s\n", output.c_str());
+	ASSERT_EQ(output.find(".io_snap"), std::string::npos);
+}
+
 class ZreplHandshakeTest : public testing::Test {
 protected:
 	/* Shared setup hook for all zrepl handshake tests - called just once */