osd: EC optimizations: add written and present shard sets to pg_log_enty_t

Add two new sets to the pg_log_entry for use by EC optimization pools.
The written shards set tracks which shards were written to, the
present shards set tracks which shards were in the acting set at the
time of the write.

An empty set (default) is used to indicate all shards. For pools without
allow_ec_optimizations the written set is empty (indicating all shards are
written) and the present set is empty and unused.

Signed-off-by: Bill Scales <bill_scales@uk.ibm.com>

diff --git a/src/osd/osd_types.cc b/src/osd/osd_types.cc
index f9da9de9e24f8315892ef4bb3f36d69fc3d24e67..366d98a6c2de6fa4c411f768e1efb034f307a12b 100644 (file)
--- a/src/osd/osd_types.cc
+++ b/src/osd/osd_types.cc
@@ -4954,7 +4954,7 @@ void pg_log_entry_t::decode_with_checksum(ceph::buffer::list::const_iterator& p)
 
 void pg_log_entry_t::encode(ceph::buffer::list &bl) const
 {
-  ENCODE_START(14, 4, bl);
+  ENCODE_START(15, 4, bl);
   encode(op, bl);
   encode(soid, bl);
   encode(version, bl);
@@ -4987,12 +4987,14 @@ void pg_log_entry_t::encode(ceph::buffer::list &bl) const
   if (op != ERROR)
     encode(return_code, bl);
   encode(op_returns, bl);
+  encode(written_shards, bl);
+  encode(present_shards, bl);
   ENCODE_FINISH(bl);
 }
 
 void pg_log_entry_t::decode(ceph::buffer::list::const_iterator &bl)
 {
-  DECODE_START_LEGACY_COMPAT_LEN(14, 4, 4, bl);
+  DECODE_START_LEGACY_COMPAT_LEN(15, 4, 4, bl);
   decode(op, bl);
   if (struct_v < 2) {
     sobject_t old_soid;
@@ -5058,6 +5060,10 @@ void pg_log_entry_t::decode(ceph::buffer::list::const_iterator &bl)
     }
     decode(op_returns, bl);
   }
+  if (struct_v >= 15) {
+    decode(written_shards, bl);
+    decode(present_shards, bl);
+  }
   DECODE_FINISH(bl);
 }
 
@@ -5107,6 +5113,8 @@ void pg_log_entry_t::dump(Formatter *f) const
       f->dump_unsigned("snap", *p);
     f->close_section();
   }
+  f->dump_stream("written_shards") << written_shards;
+  f->dump_stream("present_shards") << present_shards;
   {
     f->open_object_section("mod_desc");
     mod_desc.dump(f);

diff --git a/src/osd/osd_types.h b/src/osd/osd_types.h
index b1fb4ab14b808053a92dee123cedd59695e0c871..c1786b56d7a4c0744ec60dbd3627a908bfc1b843 100644 (file)
--- a/src/osd/osd_types.h
+++ b/src/osd/osd_types.h
@@ -4463,6 +4463,9 @@ struct pg_log_entry_t {
   bool invalid_pool; // only when decoding pool-less hobject based entries
   ObjectCleanRegions clean_regions;
 
+  shard_id_set written_shards; // EC partial writes do not update every shard
+  shard_id_set present_shards; // EC partial writes need to know set of present shards
+
   pg_log_entry_t()
    : user_version(0), return_code(0), op(0),
      invalid_hash(false), invalid_pool(false) {
@@ -4531,6 +4534,15 @@ struct pg_log_entry_t {
   }
 
   std::string get_key_name() const;
+
+  /// EC partial writes: test if a shard was written
+  bool is_written_shard(const shard_id_t shard) const {
+    return written_shards.empty() || written_shards.contains(shard);
+  }
+  bool is_present_shard(const shard_id_t shard) const {
+    return present_shards.empty() || present_shards.contains(shard);
+  }
+
   void encode_with_checksum(ceph::buffer::list& bl) const;
   void decode_with_checksum(ceph::buffer::list::const_iterator& p);