diff --git a/libraries/liblmdb/CHANGES b/libraries/liblmdb/CHANGES
index bc39d82ab5..a6304660bb 100644
--- a/libraries/liblmdb/CHANGES
+++ b/libraries/liblmdb/CHANGES
@@ -1,5 +1,12 @@
 LMDB 0.9 Change Log
 
+LMDB 0.9.17 Release Engineering
+	Fix ITS#7377 catch calloc failure
+	Fix ITS#8237 regression from ITS#7589
+	Fix ITS#8221 MDB_PAGE_FULL on delete/rebalance
+	Build
+		Create install dirs if needed (ITS#8256)
+
 LMDB 0.9.16 Release (2015/08/14)
 	Fix cursor EOF bug (ITS#8190)
 	Fix handling of subDB records (ITS#8181)
diff --git a/libraries/liblmdb/Makefile b/libraries/liblmdb/Makefile
index 2d0983eff0..62b52a9ad1 100644
--- a/libraries/liblmdb/Makefile
+++ b/libraries/liblmdb/Makefile
@@ -36,6 +36,10 @@ PROGS	= $(IPROGS) mtest mtest2 mtest3 mtest4 mtest5
 all:	$(ILIBS) $(PROGS)
 
 install: $(ILIBS) $(IPROGS) $(IHDRS)
+	mkdir -p $(DESTDIR)$(prefix)/bin
+	mkdir -p $(DESTDIR)$(prefix)/lib
+	mkdir -p $(DESTDIR)$(prefix)/include
+	mkdir -p $(DESTDIR)$(prefix)/man/man1
 	for f in $(IPROGS); do cp $$f $(DESTDIR)$(prefix)/bin; done
 	for f in $(ILIBS); do cp $$f $(DESTDIR)$(prefix)/lib; done
 	for f in $(IHDRS); do cp $$f $(DESTDIR)$(prefix)/include; done
diff --git a/libraries/liblmdb/lmdb.h b/libraries/liblmdb/lmdb.h
index 762dba2ec2..c5e5c9bf12 100644
--- a/libraries/liblmdb/lmdb.h
+++ b/libraries/liblmdb/lmdb.h
@@ -1280,7 +1280,8 @@ int  mdb_get(MDB_txn *txn, MDB_dbi dbi, MDB_val *key, MDB_val *data);
 	 *		the next update operation or the transaction ends. This saves
 	 *		an extra memcpy if the data is being generated later.
 	 *		LMDB does nothing else with this memory, the caller is expected
-	 *		to modify all of the space requested.
+	 *		to modify all of the space requested. This flag must not be
+	 *		specified if the database was opened with #MDB_DUPSORT.
 	 *	<li>#MDB_APPEND - append the given key/data pair to the end of the
 	 *		database. This option allows fast bulk loading when keys are
 	 *		already known to be in the correct order. Loading unsorted keys
@@ -1437,12 +1438,13 @@ int  mdb_cursor_get(MDB_cursor *cursor, MDB_val *key, MDB_val *data,
 	 *	<li>#MDB_RESERVE - reserve space for data of the given size, but
 	 *		don't copy the given data. Instead, return a pointer to the
 	 *		reserved space, which the caller can fill in later. This saves
-	 *		an extra memcpy if the data is being generated later.
+	 *		an extra memcpy if the data is being generated later. This flag
+	 *		must not be specified if the database was opened with #MDB_DUPSORT.
 	 *	<li>#MDB_APPEND - append the given key/data pair to the end of the
 	 *		database. No key comparisons are performed. This option allows
 	 *		fast bulk loading when keys are already known to be in the
 	 *		correct order. Loading unsorted keys with this flag will cause
-	 *		data corruption.
+	 *		a #MDB_KEYEXIST error.
 	 *	<li>#MDB_APPENDDUP - as above, but for sorted dup data.
 	 *	<li>#MDB_MULTIPLE - store multiple contiguous data elements in a
 	 *		single request. This flag may only be specified if the database
diff --git a/libraries/liblmdb/mdb.c b/libraries/liblmdb/mdb.c
index b5d59bcdb1..50d359ebb0 100644
--- a/libraries/liblmdb/mdb.c
+++ b/libraries/liblmdb/mdb.c
@@ -3553,6 +3553,9 @@ mdb_env_init_meta(MDB_env *env, MDB_meta *meta)
 	mdb_env_init_meta0(env, meta);
 
 	p = calloc(2, psize);
+	if (!p)
+		return ENOMEM;
+
 	p->mp_pgno = 0;
 	p->mp_flags = P_META;
 	*(MDB_meta *)METADATA(p) = *meta;
@@ -7682,17 +7685,23 @@ mdb_rebalance(MDB_cursor *mc)
 {
 	MDB_node	*node;
 	int rc;
-	unsigned int ptop, minkeys;
+	unsigned int ptop, minkeys, thresh;
 	MDB_cursor	mn;
 	indx_t oldki;
 
-	minkeys = 1 + (IS_BRANCH(mc->mc_pg[mc->mc_top]));
+	if (IS_BRANCH(mc->mc_pg[mc->mc_top])) {
+		minkeys = 1;
+		thresh = 1;
+	} else {
+		minkeys = 2;
+		thresh = FILL_THRESHOLD;
+	}
 	DPRINTF(("rebalancing %s page %"Z"u (has %u keys, %.1f%% full)",
 	    IS_LEAF(mc->mc_pg[mc->mc_top]) ? "leaf" : "branch",
 	    mdb_dbg_pgno(mc->mc_pg[mc->mc_top]), NUMKEYS(mc->mc_pg[mc->mc_top]),
 		(float)PAGEFILL(mc->mc_txn->mt_env, mc->mc_pg[mc->mc_top]) / 10));
 
-	if (PAGEFILL(mc->mc_txn->mt_env, mc->mc_pg[mc->mc_top]) >= FILL_THRESHOLD &&
+	if (PAGEFILL(mc->mc_txn->mt_env, mc->mc_pg[mc->mc_top]) >= thresh &&
 		NUMKEYS(mc->mc_pg[mc->mc_top]) >= minkeys) {
 		DPRINTF(("no need to rebalance page %"Z"u, above fill threshold",
 		    mdb_dbg_pgno(mc->mc_pg[mc->mc_top])));
@@ -7826,8 +7835,7 @@ mdb_rebalance(MDB_cursor *mc)
 	 * move one key from it. Otherwise we should try to merge them.
 	 * (A branch page must never have less than 2 keys.)
 	 */
-	minkeys = 1 + (IS_BRANCH(mn.mc_pg[mn.mc_top]));
-	if (PAGEFILL(mc->mc_txn->mt_env, mn.mc_pg[mn.mc_top]) >= FILL_THRESHOLD && NUMKEYS(mn.mc_pg[mn.mc_top]) > minkeys) {
+	if (PAGEFILL(mc->mc_txn->mt_env, mn.mc_pg[mn.mc_top]) >= thresh && NUMKEYS(mn.mc_pg[mn.mc_top]) > minkeys) {
 		rc = mdb_node_move(&mn, mc);
 		if (mc->mc_ki[ptop]) {
 			oldki++;
@@ -8164,7 +8172,7 @@ mdb_page_split(MDB_cursor *mc, MDB_val *newkey, MDB_val *newdata, pgno_t newpgno
 				psize = 0;
 				if (newindx <= split_indx || newindx >= nkeys) {
 					i = 0; j = 1;
-					k = newindx >= nkeys ? nkeys : split_indx+2;
+					k = newindx >= nkeys ? nkeys : split_indx+1+IS_LEAF(mp);
 				} else {
 					i = nkeys; j = -1;
 					k = split_indx-1;