From 930eae214c54f5e0e790c09f35da14dff8585ac1 Mon Sep 17 00:00:00 2001
From: Josh Durgin <jdurgin@redhat.com>
Date: Mon, 13 Feb 2017 17:42:33 -0800
Subject: [PATCH 01/11] OSDMonitor: get stripe_width via stripe_unit in ec
 profile

With bluestore, making the smallest write match min_alloc_size avoids
write amplification. With EC pools this is the stripe unit, or
stripe_width / num_data_chunks. Rather than requiring people to divide
by k to get the smallest ec write, allow it to be specified directly
via stripe_unit. Store it in the ec profile so changing a monitor
config option isn't necessary to set it.

This is particularly important for ec overwrites since they allow random i/o
which should match bluestore's checksum granularity (aka min_alloc_size).

Signed-off-by: Josh Durgin <jdurgin@redhat.com>
---
 PendingReleaseNotes                           |  7 ++-
 .../configuration/pool-pg-config-ref.rst      | 14 +++--
 doc/rados/operations/erasure-code-profile.rst | 26 ++++++--
 src/common/config.h                           |  2 -
 src/common/config_opts.h                      |  2 +-
 src/mon/OSDMonitor.cc                         | 62 +++++++++++++++----
 src/mon/OSDMonitor.h                          |  1 +
 src/test/erasure-code/TestErasureCodeLrc.cc   |  8 +--
 src/test/erasure-code/test-erasure-code.sh    |  9 ++-
 src/test/mon/osd-pool-create.sh               | 20 +++---
 10 files changed, 109 insertions(+), 42 deletions(-)

diff --git a/PendingReleaseNotes b/PendingReleaseNotes
index 3a27fdd7895..e2b043034d2 100644
--- a/PendingReleaseNotes
+++ b/PendingReleaseNotes
@@ -78,7 +78,6 @@
    still works.  See the documentation page "Mount CephFS in your
    file systems table" for details.
 
-
 12.0.1
 ------
 
@@ -106,3 +105,9 @@
 * The RGW api for getting object torrent has changed its params from 'get_torrent'
   to 'torrent' so that it can be compatible with Amazon S3. Now the request for 
   object torrent is like 'GET /ObjectName?torrent'.
+
+* The configuration option "osd pool erasure code stripe width" has
+  been replaced by "osd pool erasure code stripe unit", and given the
+  ability to be overridden by the erasure code profile setting
+  "stripe_unit". For more details see "Erasure Code Profiles" in the
+  documentation.
diff --git a/doc/rados/configuration/pool-pg-config-ref.rst b/doc/rados/configuration/pool-pg-config-ref.rst
index 9f8414128b0..9cb2c8f54ff 100644
--- a/doc/rados/configuration/pool-pg-config-ref.rst
+++ b/doc/rados/configuration/pool-pg-config-ref.rst
@@ -85,14 +85,18 @@ Ceph configuration file.
           make pool creation work in the absence of ruleset 0.
 
 
-``osd pool erasure code stripe width`` 
+``osd pool erasure code stripe unit``
 
-:Description: Sets the desired size, in bytes, of an object stripe on every
-              erasure coded pools. Every object if size S will be stored as 
-              N stripes and each stripe will be encoded/decoded individually.
+:Description: Sets the default size, in bytes, of a chunk of an object
+              stripe for erasure coded pools. Every object of size S
+              will be stored as N stripes, with each data chunk
+              receiving ``stripe unit`` bytes. Each stripe of ``N *
+              stripe unit`` bytes will be encoded/decoded
+              individually. This option can is overridden by the
+              ``stripe_unit`` setting in an erasure code profile.
 
 :Type: Unsigned 32-bit Integer
-:Default: ``4096`` 
+:Default: ``4096``
 
 
 ``osd pool default size``
diff --git a/doc/rados/operations/erasure-code-profile.rst b/doc/rados/operations/erasure-code-profile.rst
index 3262b3db8dc..ddf772d36ca 100644
--- a/doc/rados/operations/erasure-code-profile.rst
+++ b/doc/rados/operations/erasure-code-profile.rst
@@ -39,6 +39,7 @@ To create a new erasure code profile::
 	ceph osd erasure-code-profile set {name} \
              [{directory=directory}] \
              [{plugin=plugin}] \
+             [{stripe_unit=stripe_unit}] \
              [{key=value} ...] \
              [--force]
 
@@ -60,23 +61,40 @@ Where:
               plugins`_ for more information.
 
 :Type: String
-:Required: No. 
+:Required: No.
 :Default: jerasure
 
+``{stripe_unit=stripe_unit}``
+
+:Description: The amount of data in a data chunk, per stripe. For
+              example, a profile with 2 data chunks and stripe_unit=4K
+              would put the range 0-4K in chunk 0, 4K-8K in chunk 1,
+              then 8K-12K in chunk 0 again. This should be a multiple
+              of 4K for best performance. The default value is taken
+              from the monitor config option
+              ``osd_pool_erasure_code_stripe_unit`` when a pool is
+              created.  The stripe_width of a pool using this profile
+              will be the number of data chunks multiplied by this
+              stripe_unit.
+
+:Type: String
+:Required: No.
+
 ``{key=value}``
 
 :Description: The semantic of the remaining key/value pairs is defined
               by the erasure code plugin.
 
 :Type: String
-:Required: No. 
+:Required: No.
 
 ``--force``
 
-:Description: Override an existing profile by the same name.
+:Description: Override an existing profile by the same name, and allow
+              setting a non-4K-aligned stripe_unit.
 
 :Type: String
-:Required: No. 
+:Required: No.
 
 osd erasure-code-profile rm
 ============================
diff --git a/src/common/config.h b/src/common/config.h
index fe5343cdce0..bbabc14a887 100644
--- a/src/common/config.h
+++ b/src/common/config.h
@@ -38,8 +38,6 @@ enum {
 #define OSD_REP_SPLAY   1
 #define OSD_REP_CHAIN   2
 
-#define OSD_POOL_ERASURE_CODE_STRIPE_WIDTH 4096
-
 class CephContext;
 
 extern const char *CEPH_CONF_FILE_DEFAULT;
diff --git a/src/common/config_opts.h b/src/common/config_opts.h
index 2c8f947795b..836af93ebad 100644
--- a/src/common/config_opts.h
+++ b/src/common/config_opts.h
@@ -675,7 +675,7 @@ OPTION(osd_crush_update_on_start, OPT_BOOL, true)
 OPTION(osd_crush_initial_weight, OPT_DOUBLE, -1) // if >=0, the initial weight is for newly added osds.
 OPTION(osd_pool_default_crush_rule, OPT_INT, -1) // deprecated for osd_pool_default_crush_replicated_ruleset
 OPTION(osd_pool_default_crush_replicated_ruleset, OPT_INT, CEPH_DEFAULT_CRUSH_REPLICATED_RULESET)
-OPTION(osd_pool_erasure_code_stripe_width, OPT_U32, OSD_POOL_ERASURE_CODE_STRIPE_WIDTH) // in bytes
+OPTION(osd_pool_erasure_code_stripe_unit, OPT_U32, 4096) // in bytes
 OPTION(osd_pool_default_size, OPT_INT, 3)
 OPTION(osd_pool_default_min_size, OPT_INT, 0)  // 0 means no specific default; ceph will use size-size/2
 OPTION(osd_pool_default_pg_num, OPT_INT, 8) // number of PGs for new pools. Configure in global or mon section of ceph.conf
diff --git a/src/mon/OSDMonitor.cc b/src/mon/OSDMonitor.cc
index 7a5eddb1832..07a99f23724 100644
--- a/src/mon/OSDMonitor.cc
+++ b/src/mon/OSDMonitor.cc
@@ -4893,8 +4893,9 @@ void OSDMonitor::check_legacy_ec_plugin(const string& plugin, const string& prof
   }
 }
 
-int OSDMonitor::normalize_profile(const string& profilename, 
-				  ErasureCodeProfile &profile, 
+int OSDMonitor::normalize_profile(const string& profilename,
+				  ErasureCodeProfile &profile,
+				  bool force,
 				  ostream *ss)
 {
   ErasureCodeInterfaceRef erasure_code;
@@ -4904,10 +4905,39 @@ int OSDMonitor::normalize_profile(const string& profilename,
   int err = instance.factory(plugin->second,
 			     g_conf->get_val<std::string>("erasure_code_dir"),
 			     profile, &erasure_code, ss);
-  if (err)
+  if (err) {
     return err;
+  }
 
-  return erasure_code->init(profile, ss);
+  err = erasure_code->init(profile, ss);
+  if (err) {
+    return err;
+  }
+
+  auto it = profile.find("stripe_unit");
+  if (it != profile.end()) {
+    string err_str;
+    uint32_t stripe_unit = strict_si_cast<uint32_t>(it->second.c_str(), &err_str);
+    if (!err_str.empty()) {
+      *ss << "could not parse stripe_unit '" << it->second
+	  << "': " << err_str << std::endl;
+      return -EINVAL;
+    }
+    uint32_t data_chunks = erasure_code->get_data_chunk_count();
+    uint32_t chunk_size = erasure_code->get_chunk_size(stripe_unit * data_chunks);
+    if (chunk_size != stripe_unit) {
+      *ss << "stripe_unit " << stripe_unit << " does not match ec profile "
+	  << "alignment. Would be padded to " << chunk_size
+	  << std::endl;
+      return -EINVAL;
+    }
+    if ((stripe_unit % 4096) != 0 && !force) {
+      *ss << "stripe_unit should be a multiple of 4096 bytes for best performance."
+	  << "use --force to override this check" << std::endl;
+      return -EINVAL;
+    }
+  }
+  return 0;
 }
 
 int OSDMonitor::crush_ruleset_create_erasure(const string &name,
@@ -5130,12 +5160,22 @@ int OSDMonitor::prepare_pool_stripe_width(const unsigned pool_type,
     break;
   case pg_pool_t::TYPE_ERASURE:
     {
+      ErasureCodeProfile profile =
+	osdmap.get_erasure_code_profile(erasure_code_profile);
       ErasureCodeInterfaceRef erasure_code;
       err = get_erasure_code(erasure_code_profile, &erasure_code, ss);
-      uint32_t desired_stripe_width = g_conf->osd_pool_erasure_code_stripe_width;
-      if (err == 0)
-	*stripe_width = erasure_code->get_data_chunk_count() *
-	  erasure_code->get_chunk_size(desired_stripe_width);
+      if (err)
+	break;
+      uint32_t data_chunks = erasure_code->get_data_chunk_count();
+      uint32_t stripe_unit = g_conf->osd_pool_erasure_code_stripe_unit;
+      auto it = profile.find("stripe_unit");
+      if (it != profile.end()) {
+	string err_str;
+	stripe_unit = strict_si_cast<uint32_t>(it->second.c_str(), &err_str);
+	assert(err_str.empty());
+      }
+      *stripe_width = data_chunks *
+	erasure_code->get_chunk_size(stripe_unit * data_chunks);
     }
     break;
   default:
@@ -6831,14 +6871,14 @@ bool OSDMonitor::prepare_command_impl(MonOpRequestRef op,
 	if (err)
 	  goto reply;
       }
-      err = normalize_profile(name, profile_map, &ss);
+      err = normalize_profile(name, profile_map, force, &ss);
       if (err)
 	goto reply;
 
       if (osdmap.has_erasure_code_profile(name)) {
 	ErasureCodeProfile existing_profile_map =
 	  osdmap.get_erasure_code_profile(name);
-	err = normalize_profile(name, existing_profile_map, &ss);
+	err = normalize_profile(name, existing_profile_map, force, &ss);
 	if (err)
 	  goto reply;
 
@@ -6892,7 +6932,7 @@ bool OSDMonitor::prepare_command_impl(MonOpRequestRef op,
 						      &ss);
 	if (err)
 	  goto reply;
-	err = normalize_profile(name, profile_map, &ss);
+	err = normalize_profile(name, profile_map, true, &ss);
 	if (err)
 	  goto reply;
 	dout(20) << "erasure code profile set " << profile << "="
diff --git a/src/mon/OSDMonitor.h b/src/mon/OSDMonitor.h
index 32d08f78d3e..3ccbeecec76 100644
--- a/src/mon/OSDMonitor.h
+++ b/src/mon/OSDMonitor.h
@@ -297,6 +297,7 @@ private:
 			      const string& profile) const;
   int normalize_profile(const string& profilename, 
 			ErasureCodeProfile &profile,
+			bool force,
 			ostream *ss);
   int crush_ruleset_create_erasure(const string &name,
 				   const string &profile,
diff --git a/src/test/erasure-code/TestErasureCodeLrc.cc b/src/test/erasure-code/TestErasureCodeLrc.cc
index 52fce2e6b42..50543945e3a 100644
--- a/src/test/erasure-code/TestErasureCodeLrc.cc
+++ b/src/test/erasure-code/TestErasureCodeLrc.cc
@@ -615,8 +615,8 @@ TEST(ErasureCodeLrc, encode_decode)
   profile["layers"] = description_string;
   EXPECT_EQ(0, lrc.init(profile, &cerr));
   EXPECT_EQ(4U, lrc.get_data_chunk_count());
-  unsigned int stripe_width = g_conf->osd_pool_erasure_code_stripe_width;
-  unsigned int chunk_size = stripe_width / lrc.get_data_chunk_count();
+  unsigned int chunk_size = g_conf->osd_pool_erasure_code_stripe_unit;
+  unsigned int stripe_width = lrc.get_data_chunk_count() * chunk_size;
   EXPECT_EQ(chunk_size, lrc.get_chunk_size(stripe_width));
   set<int> want_to_encode;
   map<int, bufferlist> encoded;
@@ -745,8 +745,8 @@ TEST(ErasureCodeLrc, encode_decode_2)
   profile["layers"] = description_string;
   EXPECT_EQ(0, lrc.init(profile, &cerr));
   EXPECT_EQ(4U, lrc.get_data_chunk_count());
-  unsigned int stripe_width = g_conf->osd_pool_erasure_code_stripe_width;
-  unsigned int chunk_size = stripe_width / lrc.get_data_chunk_count();
+  unsigned int chunk_size = g_conf->osd_pool_erasure_code_stripe_unit;
+  unsigned int stripe_width = lrc.get_data_chunk_count() * chunk_size;
   EXPECT_EQ(chunk_size, lrc.get_chunk_size(stripe_width));
   set<int> want_to_encode;
   map<int, bufferlist> encoded;
diff --git a/src/test/erasure-code/test-erasure-code.sh b/src/test/erasure-code/test-erasure-code.sh
index b4417f664f6..cb1123febe0 100755
--- a/src/test/erasure-code/test-erasure-code.sh
+++ b/src/test/erasure-code/test-erasure-code.sh
@@ -259,8 +259,9 @@ function TEST_alignment_constraints() {
     # imposed by the stripe width
     # See http://tracker.ceph.com/issues/8622
     #
-    local stripe_width=$(ceph-conf --show-config-value osd_pool_erasure_code_stripe_width)
-    local block_size=$((stripe_width - 1))
+    local stripe_unit=$(ceph-conf --show-config-value osd_pool_erasure_code_stripe_unit)
+    eval local $(ceph osd erasure-code-profile get myprofile | grep k=)
+    local block_size=$((stripe_unit * k - 1))
     dd if=/dev/zero of=$dir/ORIGINAL bs=$block_size count=2
     rados --block-size=$block_size \
         --pool ecpool put UNALIGNED $dir/ORIGINAL || return 1
@@ -268,9 +269,7 @@ function TEST_alignment_constraints() {
 }
 
 function chunk_size() {
-    local stripe_width=$(ceph-conf --show-config-value osd_pool_erasure_code_stripe_width)
-    eval local $(ceph osd erasure-code-profile get default | grep k=)
-    echo $(($stripe_width / $k))
+    echo $(ceph-conf --show-config-value osd_pool_erasure_code_stripe_unit)
 }
 
 #
diff --git a/src/test/mon/osd-pool-create.sh b/src/test/mon/osd-pool-create.sh
index 3deef6fc10d..3fd4feef8df 100755
--- a/src/test/mon/osd-pool-create.sh
+++ b/src/test/mon/osd-pool-create.sh
@@ -122,30 +122,32 @@ function TEST_erasure_code_profile_default() {
     ceph osd erasure-code-profile ls | grep default || return 1
 }
 
-function TEST_erasure_crush_stripe_width() {
+function TEST_erasure_crush_stripe_unit() {
     local dir=$1
-    # the default stripe width is used to initialize the pool
+    # the default stripe unit is used to initialize the pool
     run_mon $dir a --public-addr $CEPH_MON
-    stripe_width=$(ceph-conf --show-config-value osd_pool_erasure_code_stripe_width)
+    stripe_unit=$(ceph-conf --show-config-value osd_pool_erasure_code_stripe_unit)
+    eval local $(ceph osd erasure-code-profile get myprofile | grep k=)
+    stripe_width = $((stripe_unit * k))
     ceph osd pool create pool_erasure 12 12 erasure
     ceph --format json osd dump | tee $dir/osd.json
     grep '"stripe_width":'$stripe_width $dir/osd.json > /dev/null || return 1
 }
 
-function TEST_erasure_crush_stripe_width_padded() {
+function TEST_erasure_crush_stripe_unit_padded() {
     local dir=$1
-    # setting osd_pool_erasure_code_stripe_width modifies the stripe_width
+    # setting osd_pool_erasure_code_stripe_unit modifies the stripe_width
     # and it is padded as required by the default plugin
     profile+=" plugin=jerasure"
     profile+=" technique=reed_sol_van"
     k=4
     profile+=" k=$k"
     profile+=" m=2"
-    expected_chunk_size=2048
-    actual_stripe_width=$(($expected_chunk_size * $k))
-    desired_stripe_width=$(($actual_stripe_width - 1))
+    actual_stripe_unit=2048
+    desired_stripe_unit=$((actual_stripe_unit - 1))
+    actual_stripe_width=$((actual_stripe_unit * k))
     run_mon $dir a \
-        --osd_pool_erasure_code_stripe_width $desired_stripe_width \
+        --osd_pool_erasure_code_stripe_unit $desired_stripe_unit \
         --osd_pool_default_erasure_code_profile "$profile" || return 1
     ceph osd pool create pool_erasure 12 12 erasure
     ceph osd dump | tee $dir/osd.json

From 6fba80c1fac7f012303b27ef1e6fdb9d90c81a40 Mon Sep 17 00:00:00 2001
From: Josh Durgin <jdurgin@redhat.com>
Date: Tue, 14 Feb 2017 00:04:12 -0800
Subject: [PATCH 02/11] osd, OSDMonitor, qa: mark ec overwrites
 non-experimental

Keep the pool flag around so we can distinguish between a pool that
should maintain hashes for each chunk, and a missing one is a bug, vs
an overwrites pool where we rely on bluestore checksums for detecting
corruption.

Signed-off-by: Josh Durgin <jdurgin@redhat.com>
---
 .../krbd/rbd-nomount/tasks/krbd_data_pool.yaml    |  6 ------
 .../ec-pool-snaps-few-objects-overwrites.yaml     |  2 +-
 .../ec-small-objects-fast-read-overwrites.yaml    |  2 +-
 .../workloads/ec-small-objects-overwrites.yaml    |  2 +-
 .../ec-snaps-few-objects-overwrites.yaml          |  2 +-
 qa/suites/rbd/cli/pool/ec-data-pool.yaml          |  4 +---
 qa/suites/rbd/librbd/pool/ec-data-pool.yaml       |  4 +---
 qa/suites/rbd/qemu/pool/ec-data-pool.yaml         |  4 +---
 qa/tasks/ceph_manager.py                          | 13 ++++++-------
 qa/tasks/rados.py                                 | 10 ++++------
 qa/workunits/rbd/krbd_data_pool.sh                |  2 +-
 src/mon/FSCommands.cc                             |  2 +-
 src/mon/MonCommands.h                             |  2 +-
 src/mon/OSDMonitor.cc                             | 15 +++++----------
 src/osd/ECBackend.cc                              | 12 ++++++------
 src/osd/osd_types.h                               |  2 +-
 16 files changed, 32 insertions(+), 52 deletions(-)

diff --git a/qa/suites/krbd/rbd-nomount/tasks/krbd_data_pool.yaml b/qa/suites/krbd/rbd-nomount/tasks/krbd_data_pool.yaml
index 24985fb79d5..35b9d67ebff 100644
--- a/qa/suites/krbd/rbd-nomount/tasks/krbd_data_pool.yaml
+++ b/qa/suites/krbd/rbd-nomount/tasks/krbd_data_pool.yaml
@@ -1,9 +1,3 @@
-overrides:
-  ceph:
-    conf:
-      global:
-        enable experimental unrecoverable data corrupting features: debug_white_box_testing_ec_overwrites
-
 tasks:
 - workunit:
     clients:
diff --git a/qa/suites/rados/thrash-erasure-code/workloads/ec-pool-snaps-few-objects-overwrites.yaml b/qa/suites/rados/thrash-erasure-code/workloads/ec-pool-snaps-few-objects-overwrites.yaml
index 7d494fdad98..d2ad70a57ec 100644
--- a/qa/suites/rados/thrash-erasure-code/workloads/ec-pool-snaps-few-objects-overwrites.yaml
+++ b/qa/suites/rados/thrash-erasure-code/workloads/ec-pool-snaps-few-objects-overwrites.yaml
@@ -12,7 +12,7 @@ tasks:
     objects: 50
     pool_snaps: true
     ec_pool: true
-    erasure_code_use_hacky_overwrites: true
+    erasure_code_use_overwrites: true
     op_weights:
       read: 100
       write: 100
diff --git a/qa/suites/rados/thrash-erasure-code/workloads/ec-small-objects-fast-read-overwrites.yaml b/qa/suites/rados/thrash-erasure-code/workloads/ec-small-objects-fast-read-overwrites.yaml
index c2e16487572..b3f831b778e 100644
--- a/qa/suites/rados/thrash-erasure-code/workloads/ec-small-objects-fast-read-overwrites.yaml
+++ b/qa/suites/rados/thrash-erasure-code/workloads/ec-small-objects-fast-read-overwrites.yaml
@@ -14,7 +14,7 @@ tasks:
     objects: 1024
     size: 16384
     ec_pool: true
-    erasure_code_use_hacky_overwrites: true
+    erasure_code_use_overwrites: true
     fast_read: true
     op_weights:
       read: 100
diff --git a/qa/suites/rados/thrash-erasure-code/workloads/ec-small-objects-overwrites.yaml b/qa/suites/rados/thrash-erasure-code/workloads/ec-small-objects-overwrites.yaml
index 80a58172169..9baacef4890 100644
--- a/qa/suites/rados/thrash-erasure-code/workloads/ec-small-objects-overwrites.yaml
+++ b/qa/suites/rados/thrash-erasure-code/workloads/ec-small-objects-overwrites.yaml
@@ -14,7 +14,7 @@ tasks:
     objects: 1024
     size: 16384
     ec_pool: true
-    erasure_code_use_hacky_overwrites: true
+    erasure_code_use_overwrites: true
     op_weights:
       read: 100
       write: 100
diff --git a/qa/suites/rados/thrash-erasure-code/workloads/ec-snaps-few-objects-overwrites.yaml b/qa/suites/rados/thrash-erasure-code/workloads/ec-snaps-few-objects-overwrites.yaml
index 00263c414cc..b7c53819938 100644
--- a/qa/suites/rados/thrash-erasure-code/workloads/ec-snaps-few-objects-overwrites.yaml
+++ b/qa/suites/rados/thrash-erasure-code/workloads/ec-snaps-few-objects-overwrites.yaml
@@ -11,7 +11,7 @@ tasks:
     ops: 4000
     objects: 50
     ec_pool: true
-    erasure_code_use_hacky_overwrites: true
+    erasure_code_use_overwrites: true
     op_weights:
       read: 100
       write: 100
diff --git a/qa/suites/rbd/cli/pool/ec-data-pool.yaml b/qa/suites/rbd/cli/pool/ec-data-pool.yaml
index a6c35e57ae3..523d7450afb 100644
--- a/qa/suites/rbd/cli/pool/ec-data-pool.yaml
+++ b/qa/suites/rbd/cli/pool/ec-data-pool.yaml
@@ -3,12 +3,10 @@ tasks:
     client.0:
       - sudo ceph osd erasure-code-profile set teuthologyprofile ruleset-failure-domain=osd m=1 k=2
       - sudo ceph osd pool create datapool 4 4 erasure teuthologyprofile
-      - sudo ceph osd pool set datapool debug_white_box_testing_ec_overwrites true
+      - sudo ceph osd pool set datapool allow_ec_overwrites true
 
 overrides:
   ceph:
     conf:
-      global:
-        enable experimental unrecoverable data corrupting features: debug_white_box_testing_ec_overwrites
       client:
         rbd default data pool: datapool
diff --git a/qa/suites/rbd/librbd/pool/ec-data-pool.yaml b/qa/suites/rbd/librbd/pool/ec-data-pool.yaml
index a6c35e57ae3..523d7450afb 100644
--- a/qa/suites/rbd/librbd/pool/ec-data-pool.yaml
+++ b/qa/suites/rbd/librbd/pool/ec-data-pool.yaml
@@ -3,12 +3,10 @@ tasks:
     client.0:
       - sudo ceph osd erasure-code-profile set teuthologyprofile ruleset-failure-domain=osd m=1 k=2
       - sudo ceph osd pool create datapool 4 4 erasure teuthologyprofile
-      - sudo ceph osd pool set datapool debug_white_box_testing_ec_overwrites true
+      - sudo ceph osd pool set datapool allow_ec_overwrites true
 
 overrides:
   ceph:
     conf:
-      global:
-        enable experimental unrecoverable data corrupting features: debug_white_box_testing_ec_overwrites
       client:
         rbd default data pool: datapool
diff --git a/qa/suites/rbd/qemu/pool/ec-data-pool.yaml b/qa/suites/rbd/qemu/pool/ec-data-pool.yaml
index a6c35e57ae3..523d7450afb 100644
--- a/qa/suites/rbd/qemu/pool/ec-data-pool.yaml
+++ b/qa/suites/rbd/qemu/pool/ec-data-pool.yaml
@@ -3,12 +3,10 @@ tasks:
     client.0:
       - sudo ceph osd erasure-code-profile set teuthologyprofile ruleset-failure-domain=osd m=1 k=2
       - sudo ceph osd pool create datapool 4 4 erasure teuthologyprofile
-      - sudo ceph osd pool set datapool debug_white_box_testing_ec_overwrites true
+      - sudo ceph osd pool set datapool allow_ec_overwrites true
 
 overrides:
   ceph:
     conf:
-      global:
-        enable experimental unrecoverable data corrupting features: debug_white_box_testing_ec_overwrites
       client:
         rbd default data pool: datapool
diff --git a/qa/tasks/ceph_manager.py b/qa/tasks/ceph_manager.py
index c6d5ef4691d..f4c9ebc7e9d 100644
--- a/qa/tasks/ceph_manager.py
+++ b/qa/tasks/ceph_manager.py
@@ -1479,7 +1479,7 @@ class CephManager:
     def create_pool_with_unique_name(self, pg_num=16,
                                      erasure_code_profile_name=None,
                                      min_size=None,
-                                     erasure_code_use_hacky_overwrites=False):
+                                     erasure_code_use_overwrites=False):
         """
         Create a pool named unique_pool_X where X is unique.
         """
@@ -1492,7 +1492,7 @@ class CephManager:
                 pg_num,
                 erasure_code_profile_name=erasure_code_profile_name,
                 min_size=min_size,
-                erasure_code_use_hacky_overwrites=erasure_code_use_hacky_overwrites)
+                erasure_code_use_overwrites=erasure_code_use_overwrites)
         return name
 
     @contextlib.contextmanager
@@ -1504,15 +1504,14 @@ class CephManager:
     def create_pool(self, pool_name, pg_num=16,
                     erasure_code_profile_name=None,
                     min_size=None,
-                    erasure_code_use_hacky_overwrites=False):
+                    erasure_code_use_overwrites=False):
         """
         Create a pool named from the pool_name parameter.
         :param pool_name: name of the pool being created.
         :param pg_num: initial number of pgs.
         :param erasure_code_profile_name: if set and !None create an
                                           erasure coded pool using the profile
-        :param erasure_code_use_hacky_overwrites: if true, use the hacky
-                                                  overwrites mode
+        :param erasure_code_use_overwrites: if true, allow overwrites
         """
         with self.lock:
             assert isinstance(pool_name, basestring)
@@ -1531,10 +1530,10 @@ class CephManager:
                     'osd', 'pool', 'set', pool_name,
                     'min_size',
                     str(min_size))
-            if erasure_code_use_hacky_overwrites:
+            if erasure_code_use_overwrites:
                 self.raw_cluster_cmd(
                     'osd', 'pool', 'set', pool_name,
-                    'debug_white_box_testing_ec_overwrites',
+                    'allow_ec_overwrites',
                     'true')
             self.pools[pool_name] = pg_num
         time.sleep(1)
diff --git a/qa/tasks/rados.py b/qa/tasks/rados.py
index 15a35e5988d..4e4f746cb05 100644
--- a/qa/tasks/rados.py
+++ b/qa/tasks/rados.py
@@ -55,9 +55,7 @@ def task(ctx, config):
               rollback: 2
               snap_remove: 0
             ec_pool: create an ec pool, defaults to False
-            erasure_code_use_hacky_overwrites: use the whitebox
-                                               testing experimental
-                                               overwrites mode
+            erasure_code_use_overwrites: test overwrites, default false
             erasure_code_profile:
               name: teuthologyprofile
               k: 2
@@ -139,7 +137,7 @@ def task(ctx, config):
         'ceph_test_rados']
     if config.get('ec_pool', False):
         args.extend(['--no-omap'])
-        if config.get('erasure_code_use_hacky_overwrites', False):
+        if config.get('erasure_code_use_overwrites', False):
             args.extend(['--no-sparse'])
         else:
             args.extend(['--ec-pool'])
@@ -230,8 +228,8 @@ def task(ctx, config):
                 else:
                     pool = manager.create_pool_with_unique_name(
                         erasure_code_profile_name=profile_name,
-                        erasure_code_use_hacky_overwrites=
-                          config.get('erasure_code_use_hacky_overwrites', False)
+                        erasure_code_use_overwrites=
+                          config.get('erasure_code_use_overwrites', False)
                     )
                     created_pools.append(pool)
                     if config.get('fast_read', False):
diff --git a/qa/workunits/rbd/krbd_data_pool.sh b/qa/workunits/rbd/krbd_data_pool.sh
index 2c83d34ca87..0b7b6c933ec 100755
--- a/qa/workunits/rbd/krbd_data_pool.sh
+++ b/qa/workunits/rbd/krbd_data_pool.sh
@@ -84,7 +84,7 @@ function count_data_objects() {
 ceph osd pool create repdata 24 24
 ceph osd erasure-code-profile set teuthologyprofile ruleset-failure-domain=osd m=1 k=2
 ceph osd pool create ecdata 24 24 erasure teuthologyprofile
-ceph osd pool set ecdata debug_white_box_testing_ec_overwrites true
+ceph osd pool set ecdata allow_ec_overwrites true
 ceph osd pool create rbdnonzero 24 24
 ceph osd pool create clonesonly 24 24
 
diff --git a/src/mon/FSCommands.cc b/src/mon/FSCommands.cc
index cc2dfe04c0b..630736c3989 100644
--- a/src/mon/FSCommands.cc
+++ b/src/mon/FSCommands.cc
@@ -849,7 +849,7 @@ int FileSystemCommandHandler::_check_pool(
          << " is an erasure-coded pool.  Use of erasure-coded pools"
          << " for CephFS metadata is not permitted";
     return -EINVAL;
-  } else if (pool->is_erasure() && !pool->is_hacky_ecoverwrites()) {
+  } else if (pool->is_erasure() && !pool->allows_ecoverwrites()) {
     // non-overwriteable EC pools are only acceptable with a cache tier overlay
     if (!pool->has_tiers() || !pool->has_read_tier() || !pool->has_write_tier()) {
       *ss << "pool '" << pool_name << "' (id '" << pool_id << "')"
diff --git a/src/mon/MonCommands.h b/src/mon/MonCommands.h
index 3945e4b3480..9506814f7be 100644
--- a/src/mon/MonCommands.h
+++ b/src/mon/MonCommands.h
@@ -737,7 +737,7 @@ COMMAND("osd pool get " \
 	"get pool parameter <var>", "osd", "r", "cli,rest")
 COMMAND("osd pool set " \
 	"name=pool,type=CephPoolname " \
-	"name=var,type=CephChoices,strings=size|min_size|crash_replay_interval|pg_num|pgp_num|crush_rule|crush_ruleset|hashpspool|nodelete|nopgchange|nosizechange|write_fadvise_dontneed|noscrub|nodeep-scrub|hit_set_type|hit_set_period|hit_set_count|hit_set_fpp|use_gmt_hitset|debug_fake_ec_pool|target_max_bytes|target_max_objects|cache_target_dirty_ratio|cache_target_dirty_high_ratio|cache_target_full_ratio|cache_min_flush_age|cache_min_evict_age|auid|min_read_recency_for_promote|min_write_recency_for_promote|fast_read|hit_set_grade_decay_rate|hit_set_search_last_n|scrub_min_interval|scrub_max_interval|deep_scrub_interval|recovery_priority|recovery_op_priority|scrub_priority|compression_mode|compression_algorithm|compression_required_ratio|compression_max_blob_size|compression_min_blob_size|csum_type|csum_min_block|csum_max_block|debug_white_box_testing_ec_overwrites " \
+	"name=var,type=CephChoices,strings=size|min_size|crash_replay_interval|pg_num|pgp_num|crush_rule|crush_ruleset|hashpspool|nodelete|nopgchange|nosizechange|write_fadvise_dontneed|noscrub|nodeep-scrub|hit_set_type|hit_set_period|hit_set_count|hit_set_fpp|use_gmt_hitset|debug_fake_ec_pool|target_max_bytes|target_max_objects|cache_target_dirty_ratio|cache_target_dirty_high_ratio|cache_target_full_ratio|cache_min_flush_age|cache_min_evict_age|auid|min_read_recency_for_promote|min_write_recency_for_promote|fast_read|hit_set_grade_decay_rate|hit_set_search_last_n|scrub_min_interval|scrub_max_interval|deep_scrub_interval|recovery_priority|recovery_op_priority|scrub_priority|compression_mode|compression_algorithm|compression_required_ratio|compression_max_blob_size|compression_min_blob_size|csum_type|csum_min_block|csum_max_block|allow_ec_overwrites " \
 	"name=val,type=CephString " \
 	"name=force,type=CephChoices,strings=--yes-i-really-mean-it,req=false", \
 	"set pool parameter <var> to <val>", "osd", "rw", "cli,rest")
diff --git a/src/mon/OSDMonitor.cc b/src/mon/OSDMonitor.cc
index 07a99f23724..5cdbac9866c 100644
--- a/src/mon/OSDMonitor.cc
+++ b/src/mon/OSDMonitor.cc
@@ -5768,18 +5768,13 @@ int OSDMonitor::prepare_command_pool_set(map<string,cmd_vartype> &cmdmap,
       ss << "expecting value 'true' or '1'";
       return -EINVAL;
     }
-  } else if (var == "debug_white_box_testing_ec_overwrites") {
+  } else if (var == "allow_ec_overwrites") {
+    if (!p.is_erasure()) {
+      ss << "ec overwrites can only be enabled for an erasure coded pool";
+      return -EINVAL;
+    }
     if (val == "true" || (interr.empty() && n == 1)) {
-      if (cct->check_experimental_feature_enabled(
-	    "debug_white_box_testing_ec_overwrites")) {
 	p.flags |= pg_pool_t::FLAG_EC_OVERWRITES;
-      } else {
-	ss << "debug_white_box_testing_ec_overwrites is an experimental feature "
-	   << "and must be enabled.  Note, this feature does not yet actually "
-	   << "work.  This flag merely enables some of the preliminary support "
-	   << "for testing purposes.";
-	return -ENOTSUP;
-      }
     } else if (val == "false" || (interr.empty() && n == 0)) {
       ss << "ec overwrites cannot be disabled once enabled";
       return -EINVAL;
diff --git a/src/osd/ECBackend.cc b/src/osd/ECBackend.cc
index ea285cc4e6f..569bb89529e 100644
--- a/src/osd/ECBackend.cc
+++ b/src/osd/ECBackend.cc
@@ -954,7 +954,7 @@ void ECBackend::handle_sub_read(
       ++i) {
     int r = 0;
     ECUtil::HashInfoRef hinfo;
-    if (!get_parent()->get_pool().is_hacky_ecoverwrites()) {
+    if (!get_parent()->get_pool().allows_ecoverwrites()) {
       hinfo = get_hash_info(i->first);
       if (!hinfo) {
 	r = -EIO;
@@ -989,7 +989,7 @@ void ECBackend::handle_sub_read(
 	  );
       }
 
-      if (!get_parent()->get_pool().is_hacky_ecoverwrites()) {
+      if (!get_parent()->get_pool().allows_ecoverwrites()) {
 	// This shows that we still need deep scrub because large enough files
 	// are read in sections, so the digest check here won't be done here.
 	// Do NOT check osd_read_eio_on_bad_digest here.  We need to report
@@ -1732,7 +1732,7 @@ bool ECBackend::try_state_to_reads()
 
   Op *op = &(waiting_state.front());
   if (op->requires_rmw() && pipeline_state.cache_invalid()) {
-    assert(get_parent()->get_pool().is_hacky_ecoverwrites());
+    assert(get_parent()->get_pool().allows_ecoverwrites());
     dout(20) << __func__ << ": blocking " << *op
 	     << " because it requires an rmw and the cache is invalid "
 	     << pipeline_state
@@ -1786,7 +1786,7 @@ bool ECBackend::try_state_to_reads()
   dout(10) << __func__ << ": " << *op << dendl;
 
   if (!op->remote_read.empty()) {
-    assert(get_parent()->get_pool().is_hacky_ecoverwrites());
+    assert(get_parent()->get_pool().allows_ecoverwrites());
     objects_read_async_no_cache(
       op->remote_read,
       [this, op](map<hobject_t,pair<int, extent_map> > &&results) {
@@ -1859,7 +1859,7 @@ bool ECBackend::try_reads_to_commit()
   dout(20) << __func__ << ": written: " << written << dendl;
   dout(20) << __func__ << ": op: " << *op << dendl;
 
-  if (!get_parent()->get_pool().is_hacky_ecoverwrites()) {
+  if (!get_parent()->get_pool().allows_ecoverwrites()) {
     for (auto &&i: op->log_entries) {
       if (i.requires_kraken()) {
 	derr << __func__ << ": log entry " << i << " requires kraken"
@@ -2365,7 +2365,7 @@ void ECBackend::be_deep_scrub(
     o.digest_present = false;
     return;
   } else {
-    if (!get_parent()->get_pool().is_hacky_ecoverwrites()) {
+    if (!get_parent()->get_pool().allows_ecoverwrites()) {
       assert(hinfo->has_chunk_hash());
       if (hinfo->get_total_chunk_size() != pos) {
 	dout(0) << "_scan_list  " << poid << " got incorrect size on read" << dendl;
diff --git a/src/osd/osd_types.h b/src/osd/osd_types.h
index 1c4e4c65a6c..550ea8dfca3 100644
--- a/src/osd/osd_types.h
+++ b/src/osd/osd_types.h
@@ -1448,7 +1448,7 @@ public:
   }
   uint64_t required_alignment() const { return stripe_width; }
 
-  bool is_hacky_ecoverwrites() const {
+  bool allows_ecoverwrites() const {
     return has_flag(FLAG_EC_OVERWRITES);
   }
 

From a219319137e8ea269c7a8ca2652b68918c1c4c15 Mon Sep 17 00:00:00 2001
From: Josh Durgin <jdurgin@redhat.com>
Date: Tue, 14 Feb 2017 17:03:19 -0800
Subject: [PATCH 03/11] qa/tasks/rados: test sparse reads with ec overwrites

Signed-off-by: Josh Durgin <jdurgin@redhat.com>
---
 qa/tasks/rados.py             | 4 +---
 qa/workunits/cephtool/test.sh | 2 +-
 2 files changed, 2 insertions(+), 4 deletions(-)

diff --git a/qa/tasks/rados.py b/qa/tasks/rados.py
index 4e4f746cb05..2ef542fd278 100644
--- a/qa/tasks/rados.py
+++ b/qa/tasks/rados.py
@@ -137,9 +137,7 @@ def task(ctx, config):
         'ceph_test_rados']
     if config.get('ec_pool', False):
         args.extend(['--no-omap'])
-        if config.get('erasure_code_use_overwrites', False):
-            args.extend(['--no-sparse'])
-        else:
+        if not config.get('erasure_code_use_overwrites', False):
             args.extend(['--ec-pool'])
     if config.get('write_fadvise_dontneed', False):
         args.extend(['--write-fadvise-dontneed'])
diff --git a/qa/workunits/cephtool/test.sh b/qa/workunits/cephtool/test.sh
index 8390c422293..94c11aeebf2 100755
--- a/qa/workunits/cephtool/test.sh
+++ b/qa/workunits/cephtool/test.sh
@@ -989,7 +989,7 @@ function test_mon_mds()
 
   # We should be permitted to use an EC pool with overwrites enabled
   # as the data pool...
-  ceph osd pool set mds-ec-pool debug_white_box_testing_ec_overwrites true --yes-i-really-mean-it
+  ceph osd pool set mds-ec-pool allow_ec_overwrites true
   ceph fs new $FS_NAME fs_metadata mds-ec-pool --force 2>$TMPFILE
   fail_all_mds $FS_NAME
   ceph fs rm $FS_NAME --yes-i-really-mean-it

From 3ca750d41dfe33c6efea4abc96d2bd426a9742b9 Mon Sep 17 00:00:00 2001
From: Josh Durgin <jdurgin@redhat.com>
Date: Thu, 16 Feb 2017 00:30:00 -0800
Subject: [PATCH 04/11] test/osd/osd-scrub-repair.sh: add ec overwrites test
 cases

Move pool and profile creation into a single function, and
add a 'allow_overwrites' parameter for it so each ec test
can be paramterized by it.

Signed-off-by: Josh Durgin <jdurgin@redhat.com>
---
 qa/workunits/ceph-helpers.sh     |  32 +++++-
 src/test/osd/osd-scrub-repair.sh | 184 +++++++++++++++++++++----------
 2 files changed, 157 insertions(+), 59 deletions(-)

diff --git a/qa/workunits/ceph-helpers.sh b/qa/workunits/ceph-helpers.sh
index 8642d376a73..c806508e204 100755
--- a/qa/workunits/ceph-helpers.sh
+++ b/qa/workunits/ceph-helpers.sh
@@ -520,6 +520,27 @@ function run_osd() {
     activate_osd $dir $id "$@"
 }
 
+function run_osd_bluestore() {
+    local dir=$1
+    shift
+    local id=$1
+    shift
+    local osd_data=$dir/$id
+
+    local ceph_disk_args
+    ceph_disk_args+=" --statedir=$dir"
+    ceph_disk_args+=" --sysconfdir=$dir"
+    ceph_disk_args+=" --prepend-to-path="
+
+    mkdir -p $osd_data
+    ceph-disk $ceph_disk_args \
+        prepare --bluestore $osd_data || return 1
+
+    local ceph_osd_args
+    ceph_osd_args+=" --enable-experimental-unrecoverable-data-corrupting-features=bluestore"
+    activate_osd $dir $id $ceph_osd_args "$@"
+}
+
 function test_run_osd() {
     local dir=$1
 
@@ -635,6 +656,7 @@ function activate_osd() {
     ceph_disk_args+=" --prepend-to-path="
 
     local ceph_args="$CEPH_ARGS"
+    ceph_args+=" --enable-experimental-unrecoverable-data-corrupting-features=bluestore"
     ceph_args+=" --osd-failsafe-full-ratio=.99"
     ceph_args+=" --osd-journal-size=100"
     ceph_args+=" --osd-scrub-load-threshold=2000"
@@ -993,10 +1015,18 @@ function objectstore_tool() {
     shift
     local osd_data=$dir/$id
 
+    local osd_type=$(cat $osd_data/type)
+
     kill_daemons $dir TERM osd.$id >&2 < /dev/null || return 1
+
+    local journal_args
+    if [ "$objectstore_type" == "filestore" ]; then
+	journal_args=" --journal-path $osd_data/journal"
+    fi
     ceph-objectstore-tool \
+	--enable-experimental-unrecoverable-data-corrupting-features=bluestore \
         --data-path $osd_data \
-        --journal-path $osd_data/journal \
+        $journal_args \
         "$@" || return 1
     activate_osd $dir $id $ceph_osd_args >&2 || return 1
     wait_for_clean >&2
diff --git a/src/test/osd/osd-scrub-repair.sh b/src/test/osd/osd-scrub-repair.sh
index 866383eb111..dc8cc500650 100755
--- a/src/test/osd/osd-scrub-repair.sh
+++ b/src/test/osd/osd-scrub-repair.sh
@@ -1,4 +1,4 @@
-#!/bin/bash
+#!/bin/bash -x
 #
 # Copyright (C) 2014 Red Hat <contact@redhat.com>
 #
@@ -154,11 +154,6 @@ function corrupt_and_repair_one() {
 function corrupt_and_repair_erasure_coded() {
     local dir=$1
     local poolname=$2
-    local profile=$3
-
-    ceph osd pool create $poolname 1 1 erasure $profile \
-        || return 1
-    wait_for_clean || return 1
 
     add_something $dir $poolname || return 1
 
@@ -176,8 +171,25 @@ function corrupt_and_repair_erasure_coded() {
 
 }
 
-function TEST_auto_repair_erasure_coded() {
+function create_ec_pool() {
+    local pool_name=$1
+    local allow_overwrites=$2
+
+    ceph osd erasure-code-profile set myprofile ruleset-failure-domain=osd $3 $4 $5 $6 $7 || return 1
+
+    ceph osd pool create "$poolname" 1 1 erasure myprofile || return 1
+
+    if [ "$allow_overwrites" = "true" ]; then
+        ceph osd pool set "$poolname" allow_ec_overwrites true || return 1
+    fi
+
+    wait_for_clean || return 1
+    return 0
+}
+
+function auto_repair_erasure_coded() {
     local dir=$1
+    local allow_overwrites=$2
     local poolname=ecpool
 
     # Launch a cluster with 5 seconds scrub interval
@@ -190,15 +202,16 @@ function TEST_auto_repair_erasure_coded() {
             --osd-scrub-min-interval=5 \
             --osd-scrub-interval-randomize-ratio=0"
     for id in $(seq 0 2) ; do
-        run_osd $dir $id $ceph_osd_args
+	if [ "$allow_overwrites" = "true" ]; then
+            run_osd_bluestore $dir $id $ceph_osd_args || return 1
+	else
+            run_osd $dir $id $ceph_osd_args || return 1
+	fi
     done
     wait_for_clean || return 1
 
     # Create an EC pool
-    ceph osd erasure-code-profile set myprofile \
-        k=2 m=1 ruleset-failure-domain=osd || return 1
-    ceph osd pool create $poolname 8 8 erasure myprofile || return 1
-    wait_for_clean || return 1
+    create_ec_pool $poolname $allow_overwrites k=2 m=1 || return 1
 
     # Put an object
     local payload=ABCDEF
@@ -222,69 +235,95 @@ function TEST_auto_repair_erasure_coded() {
     teardown $dir || return 1
 }
 
-function TEST_corrupt_and_repair_jerasure() {
+function TEST_auto_repair_erasure_coded_appends() {
+    auto_repair_erasure_coded $1 false
+}
+
+function TEST_auto_repair_erasure_coded_overwrites() {
+    auto_repair_erasure_coded $1 true
+}
+
+function corrupt_and_repair_jerasure() {
     local dir=$1
+    local allow_overwrites=$2
     local poolname=ecpool
-    local profile=myprofile
 
     setup $dir || return 1
     run_mon $dir a || return 1
     run_mgr $dir x || return 1
     for id in $(seq 0 3) ; do
-        run_osd $dir $id || return 1
+	if [ "$allow_overwrites" = "true" ]; then
+            run_osd_bluestore $dir $id || return 1
+	else
+            run_osd $dir $id || return 1
+	fi
     done
     wait_for_clean || return 1
 
-    ceph osd erasure-code-profile set $profile \
-        k=2 m=2 ruleset-failure-domain=osd || return 1
-
-    corrupt_and_repair_erasure_coded $dir $poolname $profile || return 1
+    create_ec_pool $poolname $allow_overwrites k=2 m=2 || return 1
+    corrupt_and_repair_erasure_coded $dir $poolname || return 1
 
     teardown $dir || return 1
 }
 
-function TEST_corrupt_and_repair_lrc() {
+function TEST_corrupt_and_repair_jerasure_appends() {
+    corrupt_and_repair_jerasure $1
+}
+
+function TEST_corrupt_and_repair_jerasure_overwrites() {
+    corrupt_and_repair_jerasure $1 true
+}
+
+function corrupt_and_repair_lrc() {
     local dir=$1
+    local allow_overwrites=$2
     local poolname=ecpool
-    local profile=myprofile
 
     setup $dir || return 1
     run_mon $dir a || return 1
     run_mgr $dir x || return 1
     for id in $(seq 0 9) ; do
-        run_osd $dir $id || return 1
+	if [ "$allow_overwrites" = "true" ]; then
+            run_osd_bluestore $dir $id || return 1
+	else
+            run_osd $dir $id || return 1
+	fi
     done
     wait_for_clean || return 1
 
-    ceph osd erasure-code-profile set $profile \
-        pluing=lrc \
-        k=4 m=2 l=3 \
-        ruleset-failure-domain=osd || return 1
-
-    corrupt_and_repair_erasure_coded $dir $poolname $profile || return 1
+    create_ec_pool $poolname $allow_overwrites k=4 m=2 l=3 plugin=lrc || return 1
+    corrupt_and_repair_erasure_coded $dir $poolname || return 1
 
     teardown $dir || return 1
 }
 
-function TEST_unfound_erasure_coded() {
+function TEST_corrupt_and_repair_lrc_appends() {
+    corrupt_and_repair_jerasure $1
+}
+
+function TEST_corrupt_and_repair_lrc_overwrites() {
+    corrupt_and_repair_jerasure $1 true
+}
+
+function unfound_erasure_coded() {
     local dir=$1
+    local allow_overwrites=$2
     local poolname=ecpool
     local payload=ABCDEF
 
     setup $dir || return 1
     run_mon $dir a || return 1
     run_mgr $dir x || return 1
-    run_osd $dir 0 || return 1
-    run_osd $dir 1 || return 1
-    run_osd $dir 2 || return 1
-    run_osd $dir 3 || return 1
+    for id in $(seq 0 3) ; do
+	if [ "$allow_overwrites" = "true" ]; then
+            run_osd_bluestore $dir $id || return 1
+	else
+            run_osd $dir $id || return 1
+	fi
+    done
     wait_for_clean || return 1
 
-    ceph osd erasure-code-profile set myprofile \
-      k=2 m=2 ruleset-failure-domain=osd || return 1
-    ceph osd pool create $poolname 1 1 erasure myprofile \
-      || return 1
-    wait_for_clean || return 1
+    create_ec_pool $poolname $allow_overwrites k=2 m=2 || return 1
 
     add_something $dir $poolname || return 1
 
@@ -324,27 +363,35 @@ function TEST_unfound_erasure_coded() {
     teardown $dir || return 1
 }
 
+function TEST_unfound_erasure_coded_appends() {
+    unfound_erasure_coded $1
+}
+
+function TEST_unfound_erasure_coded_overwrites() {
+    unfound_erasure_coded $1 true
+}
+
 #
 # list_missing for EC pool
 #
-function TEST_list_missing_erasure_coded() {
+function list_missing_erasure_coded() {
     local dir=$1
+    local allow_overwrites=$2
     local poolname=ecpool
-    local profile=myprofile
 
     setup $dir || return 1
     run_mon $dir a || return 1
     run_mgr $dir x || return 1
     for id in $(seq 0 2) ; do
-        run_osd $dir $id || return 1
+	if [ "$allow_overwrites" = "true" ]; then
+            run_osd_bluestore $dir $id || return 1
+	else
+            run_osd $dir $id || return 1
+	fi
     done
     wait_for_clean || return 1
 
-    ceph osd erasure-code-profile set $profile \
-        k=2 m=1 ruleset-failure-domain=osd || return 1
-    ceph osd pool create $poolname 1 1 erasure $profile \
-        || return 1
-    wait_for_clean || return 1
+    create_ec_pool $poolname $allow_overwrites k=2 m=1 || return 1
 
     # Put an object and remove the two shards (including primary)
     add_something $dir $poolname MOBJ0 || return 1
@@ -360,17 +407,17 @@ function TEST_list_missing_erasure_coded() {
     done
 
     id=${osds0[0]}
-    ceph-objectstore-tool --data-path $dir/$id --journal-path $dir/$id/journal \
+    ceph-objectstore-tool --data-path $dir/$id --enable-experimental-unrecoverable-data-corrupting-features=bluestore \
         MOBJ0 remove || return 1
     id=${osds0[1]}
-    ceph-objectstore-tool --data-path $dir/$id --journal-path $dir/$id/journal \
+    ceph-objectstore-tool --data-path $dir/$id --enable-experimental-unrecoverable-data-corrupting-features=bluestore \
         MOBJ0 remove || return 1
 
     id=${osds1[1]}
-    ceph-objectstore-tool --data-path $dir/$id --journal-path $dir/$id/journal \
+    ceph-objectstore-tool --data-path $dir/$id --enable-experimental-unrecoverable-data-corrupting-features=bluestore \
         MOBJ1 remove || return 1
     id=${osds1[2]}
-    ceph-objectstore-tool --data-path $dir/$id --journal-path $dir/$id/journal \
+    ceph-objectstore-tool --data-path $dir/$id --enable-experimental-unrecoverable-data-corrupting-features=bluestore \
         MOBJ1 remove || return 1
 
     for id in $(seq 0 2) ; do
@@ -394,6 +441,14 @@ function TEST_list_missing_erasure_coded() {
     teardown $dir || return 1
 }
 
+function TEST_list_missing_erasure_coded_appends() {
+    list_missing_erasure_coded $1 false
+}
+
+function TEST_list_missing_erasure_coded_overwrites() {
+    list_missing_erasure_coded $1 true
+}
+
 #
 # Corrupt one copy of a replicated pool
 #
@@ -1513,25 +1568,25 @@ EOF
 #
 # Test scrub errors for an erasure coded pool
 #
-function TEST_corrupt_scrub_erasure() {
+function corrupt_scrub_erasure() {
     local dir=$1
+    local allow_overwrites=$2
     local poolname=ecpool
-    local profile=myprofile
     local total_objs=5
 
     setup $dir || return 1
     run_mon $dir a || return 1
     run_mgr $dir x || return 1
     for id in $(seq 0 2) ; do
-        run_osd $dir $id || return 1
+	if [ "$allow_overwrites" = "true" ]; then
+            run_osd_bluestore $dir $id || return 1
+	else
+            run_osd $dir $id || return 1
+	fi
     done
     wait_for_clean || return 1
 
-    ceph osd erasure-code-profile set $profile \
-        k=2 m=1 ruleset-failure-domain=osd || return 1
-    ceph osd pool create $poolname 1 1 erasure $profile \
-        || return 1
-    wait_for_clean || return 1
+    create_ec_pool $poolname $allow_overwrites k=2 m=1 stripe_unit=2K --force || return 1
 
     for i in $(seq 1 $total_objs) ; do
         objname=EOBJ${i}
@@ -2151,6 +2206,11 @@ EOF
 EOF
 
     jq "$jqfilter" $dir/json | python -c "$sortkeys" | sed -e "$sedfilter" > $dir/csjson
+    if [ "$allow_overwrites" = "true" ]
+    then
+        grep -v data_digest $dir/csjson | grep -v ec_size_error > $dir/csjson
+        grep -v data_digest $dir/checkcsjson | grep -v ec_size_error > $dir/checkcsjson
+    fi
     diff -y $termwidth $dir/checkcsjson $dir/csjson || test $getjson = "yes" || return 1
     if test $getjson = "yes"
     then
@@ -2166,6 +2226,14 @@ EOF
     teardown $dir || return 1
 }
 
+function TEST_corrupt_scrub_erasure_appends() {
+    corrupt_scrub_erasure $1 false
+}
+
+function TEST_corrupt_scrub_erasure_overwrites() {
+    corrupt_scrub_erasure $1 true
+}
+
 #
 # Test to make sure that a periodic scrub won't cause deep-scrub info to be lost
 #

From 4315221179a557f9acd7527cc8be92716b15b332 Mon Sep 17 00:00:00 2001
From: Josh Durgin <jdurgin@redhat.com>
Date: Thu, 6 Apr 2017 11:40:41 -0700
Subject: [PATCH 05/11] osd,OSDMonitor: try to protect against ec overwrites
 with filestore

This isn't perfect, but it's better than nothing. Prevent enabling the
allow_ec_overwrites flag if any of a sample of pgs in the pool map to
osds using filestore. This mainly protects filestore-only clusters
from enabling it.

If a filestore osd is started later, warn in the cluster log when it
gets a pg with ec overwrites enabled.

Signed-off-by: Josh Durgin <jdurgin@redhat.com>
---
 qa/workunits/cephtool/test.sh | 12 +++++++++
 src/mon/OSDMonitor.cc         | 47 +++++++++++++++++++++++++++++++++++
 src/mon/OSDMonitor.h          |  3 +++
 src/osd/OSD.cc                |  7 ++++++
 4 files changed, 69 insertions(+)

diff --git a/qa/workunits/cephtool/test.sh b/qa/workunits/cephtool/test.sh
index 94c11aeebf2..4398ab61146 100755
--- a/qa/workunits/cephtool/test.sh
+++ b/qa/workunits/cephtool/test.sh
@@ -1320,7 +1320,19 @@ function test_mon_osd_pool()
   # should fail because the type is not the same
   expect_false ceph osd pool create replicated 12 12 erasure
   ceph osd lspools | grep replicated
+  ceph osd pool create ec_test 1 1 erasure
+  set +e
+  ceph osd metadata | grep osd_objectstore_type | grep -qc bluestore
+  if [ $? -eq 0 ]; then
+      ceph osd pool set ec_test allow_ec_overwrites true >& $TMPFILE
+      check_response $? 22 "pool must only be stored on bluestore for scrubbing to work"
+  else
+      ceph osd pool set ec_test allow_ec_overwrites true || return 1
+      expect_false ceph osd pool set ec_test allow_ec_overwrites false
+  fi
+  set -e
   ceph osd pool delete replicated replicated --yes-i-really-really-mean-it
+  ceph osd pool delete ec_test ec_test --yes-i-really-really-mean-it
 }
 
 function test_mon_osd_pool_quota()
diff --git a/src/mon/OSDMonitor.cc b/src/mon/OSDMonitor.cc
index 5cdbac9866c..1b48c8edccb 100644
--- a/src/mon/OSDMonitor.cc
+++ b/src/mon/OSDMonitor.cc
@@ -1232,6 +1232,48 @@ int OSDMonitor::load_metadata(int osd, map<string, string>& m, ostream *err)
   return 0;
 }
 
+int OSDMonitor::get_osd_objectstore_type(int osd, string *type)
+{
+  map<string, string> metadata;
+  int r = load_metadata(osd, metadata, nullptr);
+  if (r < 0)
+    return r;
+
+  auto it = metadata.find("osd_objectstore");
+  if (it == metadata.end())
+    return -ENOENT;
+  *type = it->second;
+  return 0;
+}
+
+bool OSDMonitor::is_pool_currently_all_bluestore(int64_t pool_id,
+						 const pg_pool_t &pool,
+						 ostream *err)
+{
+  // just check a few pgs for efficiency - this can't give a guarantee anyway,
+  // since filestore osds could always join the pool later
+  set<int> checked_osds;
+  for (unsigned ps = 0; ps < MIN(8, pool.get_pg_num()); ++ps) {
+    vector<int> up, acting;
+    pg_t pgid(ps, pool_id, -1);
+    osdmap.pg_to_up_acting_osds(pgid, up, acting);
+    for (int osd : up) {
+      if (checked_osds.find(osd) != checked_osds.end())
+	continue;
+      string objectstore_type;
+      int r = get_osd_objectstore_type(osd, &objectstore_type);
+      // allow with missing metadata, e.g. due to an osd never booting yet
+      if (r < 0 || objectstore_type == "bluestore") {
+	checked_osds.insert(osd);
+	continue;
+      }
+      *err << "osd." << osd << " uses " << objectstore_type;
+      return false;
+    }
+  }
+  return true;
+}
+
 int OSDMonitor::dump_osd_metadata(int osd, Formatter *f, ostream *err)
 {
   map<string,string> m;
@@ -5782,6 +5824,11 @@ int OSDMonitor::prepare_command_pool_set(map<string,cmd_vartype> &cmdmap,
       ss << "expecting value 'true', 'false', '0', or '1'";
       return -EINVAL;
     }
+    stringstream err;
+    if (!is_pool_currently_all_bluestore(pool, p, &err)) {
+      ss << "pool must only be stored on bluestore for scrubbing to work: " << err.str();
+      return -EINVAL;
+    }
   } else if (var == "target_max_objects") {
     if (interr.length()) {
       ss << "error parsing int '" << val << "': " << interr;
diff --git a/src/mon/OSDMonitor.h b/src/mon/OSDMonitor.h
index 3ccbeecec76..5144e89b6d1 100644
--- a/src/mon/OSDMonitor.h
+++ b/src/mon/OSDMonitor.h
@@ -411,6 +411,9 @@ private:
   OpTracker op_tracker;
 
   int load_metadata(int osd, map<string, string>& m, ostream *err);
+  int get_osd_objectstore_type(int osd, std::string *type);
+  bool is_pool_currently_all_bluestore(int64_t pool_id, const pg_pool_t &pool,
+				       ostream *err);
 
   // when we last received PG stats from each osd
   map<int,utime_t> last_osd_report;
diff --git a/src/osd/OSD.cc b/src/osd/OSD.cc
index 1082c66e5db..809e8f21e5d 100644
--- a/src/osd/OSD.cc
+++ b/src/osd/OSD.cc
@@ -3894,6 +3894,13 @@ int OSD::handle_pg_peering_evt(
     switch (result) {
     case RES_NONE: {
       const pg_pool_t* pp = osdmap->get_pg_pool(pgid.pool());
+      if (pp->has_flag(pg_pool_t::FLAG_EC_OVERWRITES) &&
+	  store->get_type() != "bluestore") {
+	clog->warn() << "pg " << pgid
+		     << " is at risk of silent data corruption: "
+		     << "the pool allows ec overwrites but is not stored in "
+		     << "bluestore, so deep scrubbing will not detect bitrot";
+      }
       PG::_create(*rctx.transaction, pgid, pgid.get_split_bits(pp->get_pg_num()));
       PG::_init(*rctx.transaction, pgid, pp);
 

From 25e8e1a08e8ba3b122e99f0108e12b2c120644dd Mon Sep 17 00:00:00 2001
From: Josh Durgin <jdurgin@redhat.com>
Date: Wed, 12 Apr 2017 17:59:57 -0700
Subject: [PATCH 06/11] PendingReleaseNotes: note about ec overwrites

Signed-off-by: Josh Durgin <jdurgin@redhat.com>
---
 PendingReleaseNotes | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/PendingReleaseNotes b/PendingReleaseNotes
index e2b043034d2..f010976e7ff 100644
--- a/PendingReleaseNotes
+++ b/PendingReleaseNotes
@@ -111,3 +111,8 @@
   ability to be overridden by the erasure code profile setting
   "stripe_unit". For more details see "Erasure Code Profiles" in the
   documentation.
+
+* rbd and cephfs can use erasure coding with bluestore. This may be
+  enabled by setting 'allow_ec_overwrites' to 'true' for a pool. Since
+  this relies on bluestore's checksumming to do deep scrubbing,
+  enabling this on a pool stored on filestore is not allowed.

From bad606605abbec852a261316d4e7bbe116b8c007 Mon Sep 17 00:00:00 2001
From: Josh Durgin <jdurgin@redhat.com>
Date: Wed, 12 Apr 2017 19:39:22 -0700
Subject: [PATCH 07/11] doc: update ec docs for overwrite support

Signed-off-by: Josh Durgin <jdurgin@redhat.com>
---
 doc/rados/operations/erasure-code.rst | 40 +++++++++++++++++++++------
 doc/rados/operations/pools.rst        | 11 ++++++++
 2 files changed, 42 insertions(+), 9 deletions(-)

diff --git a/doc/rados/operations/erasure-code.rst b/doc/rados/operations/erasure-code.rst
index 43a97c57caa..6878777ac72 100644
--- a/doc/rados/operations/erasure-code.rst
+++ b/doc/rados/operations/erasure-code.rst
@@ -113,12 +113,41 @@ no two *chunks* are stored in the same rack.
 More information can be found in the `erasure code profiles
 <../erasure-code-profile>`_ documentation.
 
+
+Erasure Coding with Overwrites
+------------------------------
+
+By default, erasure coded pools only work with uses like RGW that
+perform full object writes and appends.
+
+Since Luminous, partial writes for an erasure coded pool may be
+enabled with a per-pool setting. This lets RBD and Cephfs store their
+data in an erasure coded pool::
+
+    ceph osd pool set ec_pool allow_ec_overwrites true
+
+This can only be enabled on a pool residing on bluestore OSDs, since
+bluestore's checksumming is used to detect bitrot or other corruption
+during deep-scrub. In addition to being unsafe, using filestore with
+ec overwrites yields low performance compared to bluestore.
+
+Erasure coded pools do not support omap, so to use them with RBD and
+Cephfs you must instruct them to store their data in an ec pool, and
+their metadata in a replicated pool. For RBD, this means using the
+erasure coded pool as the ``--data-pool`` during image creation::
+
+    rbd create --size 1G --data-pool ec_pool replicated_pool/image_name
+
+For Cephfs, using an erasure coded pool means setting that pool in
+a `file layout<../../cephfs/file-layouts>`_.
+
+
 Erasure coded pool and cache tiering
 ------------------------------------
 
 Erasure coded pools require more resources than replicated pools and
-lack some functionalities such as partial writes. To overcome these
-limitations, it is recommended to set a `cache tier <../cache-tiering>`_
+lack some functionalities such as omap. To overcome these
+limitations, one can set up a `cache tier <../cache-tiering>`_
 before the erasure coded pool.
 
 For instance, if the pool *hot-storage* is made of fast storage::
@@ -131,13 +160,6 @@ will place the *hot-storage* pool as tier of *ecpool* in *writeback*
 mode so that every write and read to the *ecpool* are actually using
 the *hot-storage* and benefit from its flexibility and speed.
 
-It is not possible to create an RBD image on an erasure coded pool
-because it requires partial writes. It is however possible to create
-an RBD image on an erasure coded pools when a replicated pool tier set
-a cache tier::
-
-    $ rbd create --size 10G ecpool/myvolume
-
 More information can be found in the `cache tiering
 <../cache-tiering>`_ documentation.
 
diff --git a/doc/rados/operations/pools.rst b/doc/rados/operations/pools.rst
index d370677b1bb..7db41d6364a 100644
--- a/doc/rados/operations/pools.rst
+++ b/doc/rados/operations/pools.rst
@@ -301,6 +301,16 @@ You may set values for the following keys:
 :Description: The ruleset to use for mapping object placement in the cluster.
 :Type: Integer
 
+.. _allow_ec_overwrites:
+
+``allow_ec_overwrites``
+
+:Description: Whether writes to an erasure coded pool can update part
+              of an object, so cephfs and rbd can use it. See
+              `Erasure Coding with Overwrites`_ for more details.
+:Type: Boolean
+:Version: ``12.2.0`` and above
+
 .. _hashpspool:
 
 ``hashpspool``
@@ -731,3 +741,4 @@ a size of 3).
 .. _Pool, PG and CRUSH Config Reference: ../../configuration/pool-pg-config-ref
 .. _Bloom Filter: http://en.wikipedia.org/wiki/Bloom_filter
 .. _setting the number of placement groups: ../placement-groups#set-the-number-of-placement-groups
+.. _Erasure Coding with Overwrites: ../erasure-code#erasure-coding-with-overwrites

From a3e5cba3f124aa8df95816b822f15fb2a2dc79c6 Mon Sep 17 00:00:00 2001
From: Josh Durgin <jdurgin@redhat.com>
Date: Thu, 13 Apr 2017 12:04:10 -0700
Subject: [PATCH 08/11] qa: move ec overwrites tests to a bluestore-only
 subsuite

Signed-off-by: Josh Durgin <jdurgin@redhat.com>
---
 qa/suites/rados/thrash-erasure-code-overwrites/%                 | 0
 qa/suites/rados/thrash-erasure-code-overwrites/bluestore.yaml    | 1 +
 qa/suites/rados/thrash-erasure-code-overwrites/clusters          | 1 +
 qa/suites/rados/thrash-erasure-code-overwrites/fast              | 1 +
 qa/suites/rados/thrash-erasure-code-overwrites/leveldb.yaml      | 1 +
 qa/suites/rados/thrash-erasure-code-overwrites/msgr-failures     | 1 +
 qa/suites/rados/thrash-erasure-code-overwrites/rados.yaml        | 1 +
 qa/suites/rados/thrash-erasure-code-overwrites/thrashers         | 1 +
 .../workloads/ec-pool-snaps-few-objects-overwrites.yaml          | 0
 .../workloads/ec-small-objects-fast-read-overwrites.yaml         | 0
 .../workloads/ec-small-objects-overwrites.yaml                   | 0
 .../workloads/ec-snaps-few-objects-overwrites.yaml               | 0
 12 files changed, 7 insertions(+)
 create mode 100644 qa/suites/rados/thrash-erasure-code-overwrites/%
 create mode 120000 qa/suites/rados/thrash-erasure-code-overwrites/bluestore.yaml
 create mode 120000 qa/suites/rados/thrash-erasure-code-overwrites/clusters
 create mode 120000 qa/suites/rados/thrash-erasure-code-overwrites/fast
 create mode 120000 qa/suites/rados/thrash-erasure-code-overwrites/leveldb.yaml
 create mode 120000 qa/suites/rados/thrash-erasure-code-overwrites/msgr-failures
 create mode 120000 qa/suites/rados/thrash-erasure-code-overwrites/rados.yaml
 create mode 120000 qa/suites/rados/thrash-erasure-code-overwrites/thrashers
 rename qa/suites/rados/{thrash-erasure-code => thrash-erasure-code-overwrites}/workloads/ec-pool-snaps-few-objects-overwrites.yaml (100%)
 rename qa/suites/rados/{thrash-erasure-code => thrash-erasure-code-overwrites}/workloads/ec-small-objects-fast-read-overwrites.yaml (100%)
 rename qa/suites/rados/{thrash-erasure-code => thrash-erasure-code-overwrites}/workloads/ec-small-objects-overwrites.yaml (100%)
 rename qa/suites/rados/{thrash-erasure-code => thrash-erasure-code-overwrites}/workloads/ec-snaps-few-objects-overwrites.yaml (100%)

diff --git a/qa/suites/rados/thrash-erasure-code-overwrites/% b/qa/suites/rados/thrash-erasure-code-overwrites/%
new file mode 100644
index 00000000000..e69de29bb2d
diff --git a/qa/suites/rados/thrash-erasure-code-overwrites/bluestore.yaml b/qa/suites/rados/thrash-erasure-code-overwrites/bluestore.yaml
new file mode 120000
index 00000000000..1249ffda0ee
--- /dev/null
+++ b/qa/suites/rados/thrash-erasure-code-overwrites/bluestore.yaml
@@ -0,0 +1 @@
+../thrash-erasure-code/objectstore/bluestore.yaml
\ No newline at end of file
diff --git a/qa/suites/rados/thrash-erasure-code-overwrites/clusters b/qa/suites/rados/thrash-erasure-code-overwrites/clusters
new file mode 120000
index 00000000000..646ea04cd93
--- /dev/null
+++ b/qa/suites/rados/thrash-erasure-code-overwrites/clusters
@@ -0,0 +1 @@
+../thrash-erasure-code/clusters
\ No newline at end of file
diff --git a/qa/suites/rados/thrash-erasure-code-overwrites/fast b/qa/suites/rados/thrash-erasure-code-overwrites/fast
new file mode 120000
index 00000000000..6170b30e009
--- /dev/null
+++ b/qa/suites/rados/thrash-erasure-code-overwrites/fast
@@ -0,0 +1 @@
+../thrash-erasure-code/fast
\ No newline at end of file
diff --git a/qa/suites/rados/thrash-erasure-code-overwrites/leveldb.yaml b/qa/suites/rados/thrash-erasure-code-overwrites/leveldb.yaml
new file mode 120000
index 00000000000..531ecf3336a
--- /dev/null
+++ b/qa/suites/rados/thrash-erasure-code-overwrites/leveldb.yaml
@@ -0,0 +1 @@
+../thrash-erasure-code/leveldb.yaml
\ No newline at end of file
diff --git a/qa/suites/rados/thrash-erasure-code-overwrites/msgr-failures b/qa/suites/rados/thrash-erasure-code-overwrites/msgr-failures
new file mode 120000
index 00000000000..70c9ca130e2
--- /dev/null
+++ b/qa/suites/rados/thrash-erasure-code-overwrites/msgr-failures
@@ -0,0 +1 @@
+../thrash-erasure-code/msgr-failures
\ No newline at end of file
diff --git a/qa/suites/rados/thrash-erasure-code-overwrites/rados.yaml b/qa/suites/rados/thrash-erasure-code-overwrites/rados.yaml
new file mode 120000
index 00000000000..017df6f601e
--- /dev/null
+++ b/qa/suites/rados/thrash-erasure-code-overwrites/rados.yaml
@@ -0,0 +1 @@
+../thrash-erasure-code/rados.yaml
\ No newline at end of file
diff --git a/qa/suites/rados/thrash-erasure-code-overwrites/thrashers b/qa/suites/rados/thrash-erasure-code-overwrites/thrashers
new file mode 120000
index 00000000000..40ff82cf70a
--- /dev/null
+++ b/qa/suites/rados/thrash-erasure-code-overwrites/thrashers
@@ -0,0 +1 @@
+../thrash-erasure-code/thrashers
\ No newline at end of file
diff --git a/qa/suites/rados/thrash-erasure-code/workloads/ec-pool-snaps-few-objects-overwrites.yaml b/qa/suites/rados/thrash-erasure-code-overwrites/workloads/ec-pool-snaps-few-objects-overwrites.yaml
similarity index 100%
rename from qa/suites/rados/thrash-erasure-code/workloads/ec-pool-snaps-few-objects-overwrites.yaml
rename to qa/suites/rados/thrash-erasure-code-overwrites/workloads/ec-pool-snaps-few-objects-overwrites.yaml
diff --git a/qa/suites/rados/thrash-erasure-code/workloads/ec-small-objects-fast-read-overwrites.yaml b/qa/suites/rados/thrash-erasure-code-overwrites/workloads/ec-small-objects-fast-read-overwrites.yaml
similarity index 100%
rename from qa/suites/rados/thrash-erasure-code/workloads/ec-small-objects-fast-read-overwrites.yaml
rename to qa/suites/rados/thrash-erasure-code-overwrites/workloads/ec-small-objects-fast-read-overwrites.yaml
diff --git a/qa/suites/rados/thrash-erasure-code/workloads/ec-small-objects-overwrites.yaml b/qa/suites/rados/thrash-erasure-code-overwrites/workloads/ec-small-objects-overwrites.yaml
similarity index 100%
rename from qa/suites/rados/thrash-erasure-code/workloads/ec-small-objects-overwrites.yaml
rename to qa/suites/rados/thrash-erasure-code-overwrites/workloads/ec-small-objects-overwrites.yaml
diff --git a/qa/suites/rados/thrash-erasure-code/workloads/ec-snaps-few-objects-overwrites.yaml b/qa/suites/rados/thrash-erasure-code-overwrites/workloads/ec-snaps-few-objects-overwrites.yaml
similarity index 100%
rename from qa/suites/rados/thrash-erasure-code/workloads/ec-snaps-few-objects-overwrites.yaml
rename to qa/suites/rados/thrash-erasure-code-overwrites/workloads/ec-snaps-few-objects-overwrites.yaml

From 4a7c05818306f9c3f6861b9f61f526a36de0d7af Mon Sep 17 00:00:00 2001
From: Josh Durgin <jdurgin@redhat.com>
Date: Thu, 13 Apr 2017 17:31:02 -0700
Subject: [PATCH 09/11] qa: use 4k stripe_width again for test_rados_tool.sh

The stripe_unit change altered the default stripe_width to depend on k.

Signed-off-by: Josh Durgin <jdurgin@redhat.com>
---
 qa/workunits/rados/test_rados_tool.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/qa/workunits/rados/test_rados_tool.sh b/qa/workunits/rados/test_rados_tool.sh
index 3a5ddd2e859..a792835d099 100755
--- a/qa/workunits/rados/test_rados_tool.sh
+++ b/qa/workunits/rados/test_rados_tool.sh
@@ -87,7 +87,7 @@ run_expect_nosignal "$RADOS_TOOL" --object_locator "asdf" ls
 run_expect_nosignal "$RADOS_TOOL" --namespace "asdf" ls
 
 run_expect_succ "$RADOS_TOOL" mkpool "$POOL"
-run_expect_succ "$CEPH_TOOL" osd erasure-code-profile set myprofile k=2 m=1 ruleset-failure-domain=osd
+run_expect_succ "$CEPH_TOOL" osd erasure-code-profile set myprofile k=2 m=1 stripe_unit=2K ruleset-failure-domain=osd --force
 run_expect_succ "$CEPH_TOOL" osd pool create "$POOL_EC" 100 100 erasure myprofile
 
 

From f52b9d19f2bc7438afeec8d69d469fb51aeac021 Mon Sep 17 00:00:00 2001
From: Josh Durgin <jdurgin@redhat.com>
Date: Tue, 18 Apr 2017 22:54:04 -0700
Subject: [PATCH 10/11] qa: use bluestore for cephfs cephtool tests

This lets the tests enable ec overwrites

Signed-off-by: Josh Durgin <jdurgin@redhat.com>
---
 src/test/cephtool-test-mds.sh | 2 +-
 src/test/vstart_wrapper.sh    | 5 +++++
 2 files changed, 6 insertions(+), 1 deletion(-)

diff --git a/src/test/cephtool-test-mds.sh b/src/test/cephtool-test-mds.sh
index 93ea99b4f85..90f309c0427 100755
--- a/src/test/cephtool-test-mds.sh
+++ b/src/test/cephtool-test-mds.sh
@@ -18,7 +18,7 @@
 source $(dirname $0)/detect-build-env-vars.sh
 
 CEPH_CLI_TEST_DUP_COMMAND=1 \
-MDS=1 MON=1 OSD=3 MGR=0 CEPH_PORT=7200 $CEPH_ROOT/src/test/vstart_wrapper.sh \
+MDS=1 MON=1 OSD=3 MGR=0 CEPH_PORT=7200 CEPH_OBJECTSTORE="bluestore" $CEPH_ROOT/src/test/vstart_wrapper.sh \
     $CEPH_ROOT/qa/workunits/cephtool/test.sh \
     --test-mds \
     --asok-does-not-need-root
diff --git a/src/test/vstart_wrapper.sh b/src/test/vstart_wrapper.sh
index 62cd6e16b8a..17fd9836f66 100755
--- a/src/test/vstart_wrapper.sh
+++ b/src/test/vstart_wrapper.sh
@@ -30,8 +30,13 @@ function vstart_setup()
     trap "teardown $CEPH_DIR" EXIT
     export LC_ALL=C # some tests are vulnerable to i18n
     export PATH="$(pwd):${PATH}"
+    OBJSTORE_ARGS=""
+    if [ "bluestore" = "${CEPH_OBJECTSTORE}" ]; then
+        OBJSTORE_ARGS="-b"
+    fi
     $CEPH_ROOT/src/vstart.sh \
         --short \
+        $OBJSTORE_ARGS \
         -o 'paxos propose interval = 0.01' \
         -n -l || return 1
     export CEPH_CONF=$CEPH_DIR/ceph.conf

From cab2968e8f7d61f72c568a6cd6e4f08aa7777a5a Mon Sep 17 00:00:00 2001
From: Josh Durgin <jdurgin@redhat.com>
Date: Wed, 19 Apr 2017 16:46:09 -0700
Subject: [PATCH 11/11] qa: use bluestore for (k)rbd ec overwrites tests

Only bluestore supports ec overwrites

Signed-off-by: Josh Durgin <jdurgin@redhat.com>
---
 .../krbd/rbd-nomount/tasks/krbd_data_pool.yaml | 18 ++++++++++++++++++
 qa/suites/rbd/cli/pool/ec-data-pool.yaml       | 15 +++++++++++++++
 qa/suites/rbd/librbd/pool/ec-data-pool.yaml    | 15 +++++++++++++++
 qa/suites/rbd/qemu/pool/ec-data-pool.yaml      | 15 +++++++++++++++
 4 files changed, 63 insertions(+)

diff --git a/qa/suites/krbd/rbd-nomount/tasks/krbd_data_pool.yaml b/qa/suites/krbd/rbd-nomount/tasks/krbd_data_pool.yaml
index 35b9d67ebff..1dab39755c8 100644
--- a/qa/suites/krbd/rbd-nomount/tasks/krbd_data_pool.yaml
+++ b/qa/suites/krbd/rbd-nomount/tasks/krbd_data_pool.yaml
@@ -1,3 +1,21 @@
+overrides:
+  thrashosds:
+    bdev_inject_crash: 2
+    bdev_inject_crash_probability: .5
+  ceph:
+    fs: xfs
+    conf:
+      osd: # force bluestore since it's required for ec overwrites
+        osd objectstore: bluestore
+        bluestore block size: 96636764160
+        debug bluestore: 30
+        debug bdev: 20
+        debug bluefs: 20
+        debug rocksdb: 10
+        enable experimental unrecoverable data corrupting features: "*"
+        osd debug randomize hobject sort order: false
+# this doesn't work with failures bc the log writes are not atomic across the two backends
+#        bluestore bluefs env mirror: true
 tasks:
 - workunit:
     clients:
diff --git a/qa/suites/rbd/cli/pool/ec-data-pool.yaml b/qa/suites/rbd/cli/pool/ec-data-pool.yaml
index 523d7450afb..9558cc63a8c 100644
--- a/qa/suites/rbd/cli/pool/ec-data-pool.yaml
+++ b/qa/suites/rbd/cli/pool/ec-data-pool.yaml
@@ -6,7 +6,22 @@ tasks:
       - sudo ceph osd pool set datapool allow_ec_overwrites true
 
 overrides:
+  thrashosds:
+    bdev_inject_crash: 2
+    bdev_inject_crash_probability: .5
   ceph:
+    fs: xfs
     conf:
       client:
         rbd default data pool: datapool
+      osd: # force bluestore since it's required for ec overwrites
+        osd objectstore: bluestore
+        bluestore block size: 96636764160
+        debug bluestore: 30
+        debug bdev: 20
+        debug bluefs: 20
+        debug rocksdb: 10
+        enable experimental unrecoverable data corrupting features: "*"
+        osd debug randomize hobject sort order: false
+# this doesn't work with failures bc the log writes are not atomic across the two backends
+#        bluestore bluefs env mirror: true
diff --git a/qa/suites/rbd/librbd/pool/ec-data-pool.yaml b/qa/suites/rbd/librbd/pool/ec-data-pool.yaml
index 523d7450afb..9558cc63a8c 100644
--- a/qa/suites/rbd/librbd/pool/ec-data-pool.yaml
+++ b/qa/suites/rbd/librbd/pool/ec-data-pool.yaml
@@ -6,7 +6,22 @@ tasks:
       - sudo ceph osd pool set datapool allow_ec_overwrites true
 
 overrides:
+  thrashosds:
+    bdev_inject_crash: 2
+    bdev_inject_crash_probability: .5
   ceph:
+    fs: xfs
     conf:
       client:
         rbd default data pool: datapool
+      osd: # force bluestore since it's required for ec overwrites
+        osd objectstore: bluestore
+        bluestore block size: 96636764160
+        debug bluestore: 30
+        debug bdev: 20
+        debug bluefs: 20
+        debug rocksdb: 10
+        enable experimental unrecoverable data corrupting features: "*"
+        osd debug randomize hobject sort order: false
+# this doesn't work with failures bc the log writes are not atomic across the two backends
+#        bluestore bluefs env mirror: true
diff --git a/qa/suites/rbd/qemu/pool/ec-data-pool.yaml b/qa/suites/rbd/qemu/pool/ec-data-pool.yaml
index 523d7450afb..9558cc63a8c 100644
--- a/qa/suites/rbd/qemu/pool/ec-data-pool.yaml
+++ b/qa/suites/rbd/qemu/pool/ec-data-pool.yaml
@@ -6,7 +6,22 @@ tasks:
       - sudo ceph osd pool set datapool allow_ec_overwrites true
 
 overrides:
+  thrashosds:
+    bdev_inject_crash: 2
+    bdev_inject_crash_probability: .5
   ceph:
+    fs: xfs
     conf:
       client:
         rbd default data pool: datapool
+      osd: # force bluestore since it's required for ec overwrites
+        osd objectstore: bluestore
+        bluestore block size: 96636764160
+        debug bluestore: 30
+        debug bdev: 20
+        debug bluefs: 20
+        debug rocksdb: 10
+        enable experimental unrecoverable data corrupting features: "*"
+        osd debug randomize hobject sort order: false
+# this doesn't work with failures bc the log writes are not atomic across the two backends
+#        bluestore bluefs env mirror: true