librbd: add SnapshotNamespace to ImageCtx #12970

VictorDenisov · 2017-01-17T22:55:22Z

This PR is for preliminary review. PR #11544 is supposed to be rebased on top of this PR.

dillaman · 2017-01-18T00:04:54Z

src/librbd/ImageCtx.cc

  {
    assert(snap_lock.is_locked());
-    map<string, snap_t>::const_iterator it =
-      snap_ids.find(in_snap_name);
+    map<pair<cls::rbd::SnapshotNamespace, string>, snap_t>::const_iterator it =


Nit: feel free to use auto here to avoid having very long typedef types

dillaman · 2017-01-18T00:05:36Z

src/librbd/ImageCtx.cc

-    map<string, snap_t>::const_iterator it =
-      snap_ids.find(in_snap_name);
+    map<pair<cls::rbd::SnapshotNamespace, string>, snap_t>::const_iterator it =
+      snap_ids.find(make_pair(in_snap_namespace, in_snap_name));


Nit: not a big deal, but you can also just use {in_snap_namespace, in_snap_name} to auto-create the pair.

dillaman · 2017-01-18T00:06:42Z

src/librbd/ImageCtx.cc

@@ -552,15 +555,19 @@ struct C_InvalidateCache : public Context {
    SnapInfo info(in_snap_name, in_snap_namespace,
 		  in_size, parent, protection_status, flags);
    snap_info.insert(pair<snap_t, SnapInfo>(id, info));
-    snap_ids.insert(pair<string, snap_t>(in_snap_name, id));
+    snap_ids.insert(pair<pair<cls::rbd::SnapshotNamespace, string>,


Nit: good example of using C++11 brace initialization to make the code look cleaner

VictorDenisov · 2017-01-18T17:50:27Z

@dillaman Could you please take a look at internal.cc in this PR request. Basically every method that was supposed to look up snapshots by their names now receives an additional argument - snapshot namespace. I presume CLI code will be using UserSnapshotNamespace by default. As I understand we don't want to add more parameters to command line tool in order to allow users to specify what snap shot namespace they are using. Please clarify if I'm moving into the right direction.

dillaman · 2017-01-18T18:25:15Z

src/librbd/ImageCtx.cc

  {
    assert(snap_lock.is_wlocked());
    snaps.erase(std::remove(snaps.begin(), snaps.end(), id), snaps.end());
    snap_info.erase(id);
-    snap_ids.erase(in_snap_name);
+    snap_ids.erase(make_pair(in_snap_namespace, in_snap_name));


Nit: feel free to use brace initialization here as well

dillaman · 2017-01-18T18:25:43Z

src/librbd/ImageCtx.cc

@@ -428,11 +430,11 @@ struct C_InvalidateCache : public Context {
    data_ctx.snap_set_read(snap_id);


~~snap_namespace = {};~~

dillaman · 2017-01-18T18:26:46Z

src/librbd/DiffIterate.cc

@@ -268,7 +268,8 @@ int DiffIterate::execute() {
    r = 0;
    if (m_image_ctx.parent && overlap > 0) {
      ldout(cct, 10) << " first getting parent diff" << dendl;
-      DiffIterate diff_parent(*m_image_ctx.parent, NULL, 0, overlap,
+      DiffIterate diff_parent(*m_image_ctx.parent, cls::rbd::UnknownSnapshotNamespace(),


Nit: replace cls::rbd::UnknownSnapshotNamespace with {}

I presume in order to use this I should reorder Namespaces in the definition of Variant. So that UnknownSnapshotNamespace is the first one.

If it compiles, it doesn't really matter since the name is null

dillaman · 2017-01-18T18:27:13Z

src/librbd/DiffIterate.cc

@@ -268,7 +268,8 @@ int DiffIterate::execute() {
    r = 0;
    if (m_image_ctx.parent && overlap > 0) {
      ldout(cct, 10) << " first getting parent diff" << dendl;
-      DiffIterate diff_parent(*m_image_ctx.parent, NULL, 0, overlap,
+      DiffIterate diff_parent(*m_image_ctx.parent, cls::rbd::UnknownSnapshotNamespace(),
+			      NULL, 0, overlap,


Nit: while you are here, can you chance NULL to nullptr

dillaman · 2017-01-18T18:33:58Z

src/librbd/internal.cc

@@ -677,13 +677,14 @@ void filter_out_mirror_watchers(ImageCtx *ictx,
    return 0;
  }

-  int flatten_children(ImageCtx *ictx, const char* snap_name, ProgressContext& pctx)
+  int flatten_children(ImageCtx *ictx, const cls::rbd::SnapshotNamespace& snap_namespace,


You are safe to assume this is for an ImageSnapshotNamespace only

dillaman · 2017-01-18T18:34:06Z

src/librbd/internal.cc

-
-  int snap_is_protected(ImageCtx *ictx, const char *snap_name,
-			bool *is_protected)
+  int snap_is_protected(ImageCtx *ictx, const cls::rbd::SnapshotNamespace& snap_namespace,


You are safe to assume this is for an ImageSnapshotNamespace only

dillaman · 2017-01-18T18:34:14Z

src/librbd/internal.cc

    return 0;
  }

-  int snap_remove(ImageCtx *ictx, const char *snap_name, uint32_t flags, ProgressContext& pctx)
+  int snap_remove(ImageCtx *ictx, const cls::rbd::SnapshotNamespace& snap_namespace,


You are safe to assume this is for an ImageSnapshotNamespace only

dillaman · 2017-01-18T23:18:27Z

src/librbd/internal.h

@@ -184,7 +182,8 @@ namespace librbd {
  int64_t read_iterate(ImageCtx *ictx, uint64_t off, uint64_t len,
 		       int (*cb)(uint64_t, size_t, const char *, void *),
 		       void *arg);
-  int diff_iterate(ImageCtx *ictx, const char *fromsnapname, uint64_t off,
+  int diff_iterate(ImageCtx *ictx, const cls::rbd::SnapshotNamespace& fromsnapnamespace,


Nit: can you add underscores between words: fromsnapnamespace -> from_snap_namespace

dillaman · 2017-01-18T23:18:47Z

src/librbd/internal.cc

@@ -2343,7 +2317,8 @@ void filter_out_mirror_watchers(ImageCtx *ictx,
    return total_read;
  }

-  int diff_iterate(ImageCtx *ictx, const char *fromsnapname, uint64_t off,
+  int diff_iterate(ImageCtx *ictx, const cls::rbd::SnapshotNamespace fromsnapnamespace,


Nit: same underscore comment here

VictorDenisov · 2017-01-21T02:17:26Z

@dillaman Could you please take a look at these changes. So far I changed all required files in librbd directory, unit tests are still to be fixed.

dillaman · 2017-01-23T14:43:09Z

src/librbd/ImageState.h

@@ -73,6 +74,7 @@ class ImageState {
    ActionType action_type;
    uint64_t refresh_seq = 0;
    std::string snap_name;
+    cls::rbd::SnapshotNamespace snap_namespace;


You need to update ImageState::set_snap to take a snapshot namespace and assign the snapshot namespace to this param.

dillaman · 2017-01-23T14:44:50Z

src/librbd/WatchNotifyTypes.h


  void encode(bufferlist &bl) const;
  void decode(__u8 version, bufferlist::iterator &iter);
  void dump(Formatter *f) const;

 protected:
  SnapPayloadBase() {}
-  SnapPayloadBase(const std::string &name) : snap_name(name) {}
+  SnapPayloadBase(const std::string &name, const cls::rbd::SnapshotNamespace& _snap_namespace) : snap_name(name), snap_namespace(_snap_namespace) {}


Nit: avoid going past 80 characters in a single line. This comment applies to all changes in this file.

dillaman · 2017-01-23T14:46:57Z

src/librbd/image/RefreshParentRequest.cc

@@ -165,11 +165,17 @@ void RefreshParentRequest<I>::send_set_parent_snap() {
    return;
  }

+  cls::rbd::SnapshotNamespace snap_namespace;


Merge this logic w/ the logic on line 159. This implies we need a new helper method ImageCtx::get_snap which will populate both the snap name and namespace. This avoids weird errors where you drop the lock and the snapshot disappears. Also fixes the case where the potential error isn't currently handled below.

dillaman · 2017-01-23T14:47:14Z

src/librbd/image/SetSnapRequest.cc

                                  Context *on_finish)
-  : m_image_ctx(image_ctx), m_snap_name(snap_name), m_on_finish(on_finish),
+  : m_image_ctx(image_ctx), m_snap_name(snap_name), m_snap_namespace(snap_namespace), m_on_finish(on_finish),


Nit: max line length

dillaman · 2017-01-23T14:47:38Z

src/librbd/internal.cc

-  int diff_iterate(ImageCtx *ictx, const cls::rbd::SnapshotNamespace from_snap_namespace,
-		   const char *fromsnapname, uint64_t off,
+  int diff_iterate(ImageCtx *ictx, const char *fromsnapname,
+		   const cls::rbd::SnapshotNamespace from_snap_namespace, uint64_t off,


Nit: pass by ref

dillaman · 2017-01-23T14:48:49Z

src/librbd/journal/Types.h


 protected:
  SnapEventBase() {
  }
-  SnapEventBase(uint64_t op_tid, const std::string &_snap_name)
-    : OpEventBase(op_tid), snap_name(_snap_name) {
+  SnapEventBase(uint64_t op_tid, const std::string &_snap_name, const cls::rbd::SnapshotNamespace& _snap_namespace)


Nit: max line length throughout the changes to this file

dillaman · 2017-01-23T14:51:30Z

src/librbd/journal/Types.h

@@ -121,12 +121,13 @@ struct OpFinishEvent : public OpEventBase {

 struct SnapEventBase : public OpEventBase {
  std::string snap_name;
+  cls::rbd::SnapshotNamespace snap_namespace;


Only snap create and remove would require the namespace -- all other ops imply the user snapshot namespace. Either create a new SnapNamespaceBase for those two structs to inherit from or just add the snap namespace to those two ops. This comment applies to the notifications in WatchNotifyTypes as well.

... also, I don't see the namespace being encoded/decoded. Ensure that you bump the struct version when you do add that.

I thought that only user facing functions assume UserSnapshotNamespace by default. Internally we may need to flatten image snapshots, protect and unprotect them. For example when we clone group snapshots into another consistency group). As far as I understand these types are required for those internal operations. Am I wrong?

So, my impression that as maximum - rename op doesn't need this field.

Fair enough -- although I'm still not sure how cloning is going to work for groups since OpenStack allows you to remove images from the group and yet still create group clones from a group snapshot. Lots of holes in the design.

Per my conversation with Xing: http://git.net/ml/openstack-dev/2016-12/msg00323.html
When we clone a consistency group we basically create a new image from every snapshot in this group snapshot. Then we make a group of those images.

Yes, but the creation of new images should be a clone operation (that's RBD's core design win for OpenStack). That implies that there will be lots of edge cases to handle OpenStack since OpenStack won't treat the images as dependent. We can tackle that issue later.

dillaman · 2017-01-23T14:55:19Z

src/librbd/image/RefreshRequest.cc

@@ -1008,7 +1008,7 @@ void RefreshRequest<I>::apply() {
    m_image_ctx.snapc = m_snapc;

    if (m_image_ctx.snap_id != CEPH_NOSNAP &&
-        m_image_ctx.get_snap_id(m_image_ctx.snap_name) != m_image_ctx.snap_id) {
+        m_image_ctx.get_snap_id(m_image_ctx.snap_namespace, m_image_ctx.snap_name) != m_image_ctx.snap_id) {


Nit: line length

dillaman · 2017-01-26T13:52:44Z

@VictorDenisov The PR doesn't compile right now

VictorDenisov · 2017-01-26T17:26:40Z

yeah, I'm working on compiling unittest_rbd_mirror target. I hope it's the last one.

VictorDenisov · 2017-01-27T01:52:40Z

Now I need to fix the tests and it's done.

VictorDenisov · 2017-01-30T23:28:07Z

@dillaman done

dillaman · 2017-01-31T00:55:17Z

src/tools/rbd_mirror/image_sync/SnapshotCopyRequest.cc

@@ -247,7 +248,7 @@ void SnapshotCopyRequest<I>::send_snap_remove() {
    SnapshotCopyRequest<I>, &SnapshotCopyRequest<I>::handle_snap_remove>(
      this);
  RWLock::RLocker owner_locker(m_local_image_ctx->owner_lock);
-  m_local_image_ctx->operations->execute_snap_remove(m_snap_name.c_str(), ctx);
+  m_local_image_ctx->operations->execute_snap_remove(m_snap_name.c_str(), cls::rbd::UserSnapshotNamespace(), ctx);


Nit: 80 character line limit

dillaman · 2017-01-31T00:56:15Z

src/tools/rbd_mirror/image_sync/SnapshotCopyRequest.cc

@@ -27,11 +27,11 @@ const std::string &get_snapshot_name(I *image_ctx, librados::snap_t snap_id) {
  auto snap_it = std::find_if(image_ctx->snap_ids.begin(),
                              image_ctx->snap_ids.end(),
                              [snap_id](
-      const std::pair<std::string, librados::snap_t> &pair) {
+      const std::pair<std::pair<cls::rbd::SnapshotNamespace, std::string>, librados::snap_t> &pair) {


Nit: perhaps just switch to an auto & here to avoid the long line

It looks like auto in lambdas produces a warning unless I specify parameter -std=c++1y. Do you prefer the warning or just wrap the lines?

It's fine to leave it

Sorry, by "leave it" you mean wrap the lines. Correct?

No warnings, no switching to C++14 standard -- just try to avoid >80 characters if possible

dillaman · 2017-01-31T00:58:23Z

src/librbd/ImageCtx.cc

  {
    assert(snap_lock.is_wlocked());
-    snap_t in_snap_id = get_snap_id(in_snap_name);
+    snap_t in_snap_id = get_snap_id(in_snap_name, in_snap_namespace);


Nit: still missing a consistent calling convention for snapshot namespace + name. I would just suggest always using namespace followed by name as the convention and modify where appropriate since that follows the map key.

Is it ok to have name and then namespace? I think this is the only place where I confused this order. I was going for name, namespace order.

That's fine -- just want to keep it consistent. The rationale for <namespace, name> is because that is how it's actually stored in ImageCtx right now (but it could also be switched to <name, namespace>)

It seems like I can't change the order of namespace and name in snap_ids map. I use function: lower_bound when I look for a snapshot by its namespace when I don't know its name. Therefore namespace should go first. I'll need to change everything to be in the order: name, namespace. Unless you can change my view as you often do)

dillaman · 2017-01-31T01:08:20Z

src/librbd/ImageWatcher.cc

  assert(m_image_ctx.owner_lock.is_locked());
  assert(m_image_ctx.exclusive_lock &&
         !m_image_ctx.exclusive_lock->is_lock_owner());

-  notify_lock_owner(SnapRenamePayload(src_snap_id, dst_snap_name), on_finish);
+  notify_lock_owner(SnapRenamePayload(src_snap_id, dst_snap_name, cls::rbd::UserSnapshotNamespace()), on_finish);


Nit: line length

dillaman · 2017-01-31T01:10:41Z

src/librbd/journal/Types.h

@@ -173,7 +178,9 @@ struct SnapRenameEvent : public SnapEventBase {
  SnapRenameEvent(uint64_t op_tid, uint64_t src_snap_id,
                  const std::string &src_snap_name,
                  const std::string &dest_snap_name)
-    : SnapEventBase(op_tid, dest_snap_name), snap_id(src_snap_id),
+    : SnapEventBase(op_tid, dest_snap_name,
+      cls::rbd::UserSnapshotNamespace()), // Only snapshots from user namespace are allowed to be renamed.


Nit: indentation

dillaman · 2017-01-31T01:25:14Z

src/test/rbd_mirror/image_sync/test_mock_ImageCopyRequest.cc

-    EXPECT_CALL(mock_image_ctx, get_snap_id(_))
-      .WillRepeatedly(Invoke([&mock_image_ctx](std::string snap_name) {
+    EXPECT_CALL(mock_image_ctx, get_snap_id(_, _))
+      .WillRepeatedly(Invoke([&mock_image_ctx](std::string snap_name, cls::rbd::SnapshotNamespace snap_namespace) {


Nit: line length

dillaman · 2017-01-31T01:25:19Z

src/test/rbd_mirror/image_sync/test_mock_ImageCopyRequest.cc

@@ -313,7 +317,7 @@ TEST_F(TestMockImageSyncImageCopyRequest, SnapshotSubset) {
  ASSERT_EQ(0, create_snap("snap1"));
  ASSERT_EQ(0, create_snap("snap2"));
  ASSERT_EQ(0, create_snap("snap3"));
-  m_client_meta.sync_points = {{"snap3", "snap2", boost::none}};
+  m_client_meta.sync_points = {{"snap3", "snap2", boost::none, cls::rbd::UserSnapshotNamespace()}};


Nit: line length

dillaman · 2017-01-31T01:25:25Z

src/test/rbd_mirror/image_sync/test_mock_ImageCopyRequest.cc

@@ -383,7 +387,7 @@ TEST_F(TestMockImageSyncImageCopyRequest, RestartCatchup) {

 TEST_F(TestMockImageSyncImageCopyRequest, RestartPartialSync) {
  ASSERT_EQ(0, create_snap("snap1"));
-  m_client_meta.sync_points = {{"snap1", librbd::journal::MirrorPeerSyncPoint::ObjectNumber{0U}}};
+  m_client_meta.sync_points = {{"snap1", librbd::journal::MirrorPeerSyncPoint::ObjectNumber{0U}, cls::rbd::UserSnapshotNamespace()}};


Nit: line length

dillaman · 2017-01-31T01:25:51Z

src/test/rbd_mirror/test_PoolWatcher.cc

@@ -131,7 +131,7 @@ TestPoolWatcher() : m_lock("TestPoolWatcherLock"),
      ictx->state->open(false);
      EXPECT_EQ(0, ictx->operations->snap_create(snap_name.c_str(),
 						 cls::rbd::UserSnapshotNamespace()));
-      EXPECT_EQ(0, ictx->operations->snap_protect(snap_name.c_str()));
+      EXPECT_EQ(0, ictx->operations->snap_protect(snap_name.c_str(), cls::rbd::UserSnapshotNamespace()));


Nit: line length

dillaman · 2017-01-31T01:26:00Z

src/test/rbd_mirror/test_mock_ImageSync.cc

@@ -197,8 +197,8 @@ class TestMockImageSync : public TestMockFixture {
    EXPECT_CALL(mock_sync_point_create_request, send())
      .WillOnce(Invoke([this, &mock_local_image_ctx, &mock_sync_point_create_request, r]() {
          if (r == 0) {
-            mock_local_image_ctx.snap_ids["snap1"] = 123;
-            m_client_meta.sync_points.emplace_back("snap1", boost::none);
+            mock_local_image_ctx.snap_ids[{cls::rbd::UserSnapshotNamespace(), "snap1"}] = 123;


Nit: line length

dillaman · 2017-01-31T01:27:07Z

@VictorDenisov I would recommend just squashing these three commits into a single commit since they don't compile independently.

VictorDenisov · 2017-03-06T09:25:52Z

@dillaman it's ready

dillaman · 2017-03-08T17:34:37Z

retest this please

dillaman · 2017-03-09T01:40:13Z

src/librbd/ImageCtx.cc

@@ -435,11 +437,11 @@ struct C_InvalidateCache : public Context {
    data_ctx.snap_set_read(snap_id);


~~Nit: snap_namespace = {};~~

dillaman · 2017-03-09T01:45:12Z

src/librbd/WatchNotifyTypes.h

@@ -218,6 +218,7 @@ struct SnapPayloadBase {
 public:
  static const bool CHECK_FOR_REFRESH = true;

+  cls::rbd::SnapshotNamespace snap_namespace;


This isn't encoded/decoded/dumped (in the respective methods). It's in SnapCreatePayload -- you can probably do some magic where if the version == 5, you decode it in SnapCreatePayload -- otherwise let the base class handle it.

dillaman · 2017-03-09T01:45:47Z

src/librbd/WatchNotifyTypes.h

-  SnapProtectPayload(const std::string &name) : SnapPayloadBase(name) {}
+  SnapProtectPayload(const cls::rbd::SnapshotNamespace& snap_namespace,
+		     const std::string &name)
+			    : SnapPayloadBase(snap_namespace, name) {}


Nit: reduce indentation

dillaman · 2017-03-09T01:51:39Z

src/librbd/journal/Types.h

@@ -140,13 +140,17 @@ struct OpFinishEvent : public OpEventBase {
 };

 struct SnapEventBase : public OpEventBase {
+  cls::rbd::SnapshotNamespace snap_namespace;


Need to encode/decode/dump -- since SnapCreateEvent already expects it at version >= 3 && < 6, it should decode it directly. Otherwise, if version >= 6, let the base class decode it (and encode it unconditionally). Also -- bump the version to 6.

dillaman

Few more minor nits (and version encoding/decoding issues) and it should be ready for testing

dillaman · 2017-03-13T18:51:09Z

@VictorDenisov rebase is required

VictorDenisov · 2017-03-15T14:51:39Z

@dillaman ready

dillaman · 2017-03-15T15:58:15Z

src/librbd/journal/Types.cc

  boost::apply_visitor(EncodeVisitor(bl), client_meta);
  ENCODE_FINISH(bl);
 }

 void ClientData::decode(bufferlist::iterator& it) {
-  DECODE_START(1, it);
+  DECODE_START(2, it);


Nit: this shouldn't change -- it should remain at version 1 to indicate we can successfully still decode version 1 structures

dillaman · 2017-03-15T16:04:20Z

src/librbd/journal/Types.h

@@ -194,8 +199,10 @@ struct SnapRenameEvent : public SnapEventBase {
  SnapRenameEvent(uint64_t op_tid, uint64_t src_snap_id,
                  const std::string &src_snap_name,
                  const std::string &dest_snap_name)
-    : SnapEventBase(op_tid, dest_snap_name), snap_id(src_snap_id),
-      src_snap_name(src_snap_name) {
+    : SnapEventBase(op_tid, cls::rbd::UserSnapshotNamespace(),


Since SnapRenameEvent encodes a snap id and source snap name, it can no longer inherit from SnapEventBase since SnapEventBase encode function will inject new parameters into the struct between SnaRenameEvent's parameters.

The essence of this huge change is to have two keys for indexing snap ids in ImageCtx. It used to be a map of (snap_name -> snap_id) now it's (snap_namespace, snap_name) -> snap_id. Therefore now snapshots can have the same name if they are in different namespaces. All the remaining changes are a consequece of this change of ImageCtx.snap_ids field. The only exception is: we assume that you can't rename snapshots from GroupSnapshotNamespaces. So rename operation always assumes UserSnapshotNamespace. Signed-off-by: Victor Denisov <denisovenator@gmail.com>

VictorDenisov · 2017-03-16T16:51:43Z

@dillaman i dropped the inheritance from SnapEventBase for SnapRenameEvent.

dillaman

lgtm

dillaman · 2017-03-17T16:59:59Z

Manually merged to master branch

librbd: add SnapshotNamespace to ImageCtx Reviewed-by: Jason Dillaman <dillaman@redhat.com>

VictorDenisov mentioned this pull request Jan 17, 2017

librbd: group snapshots #11544

Closed

liewegas added feature rbd labels Jan 17, 2017

dillaman changed the title ~~DoNotMerge RBD:Add SnapshotNamespace to ImageCtx methods~~ [DNM] librbd: add SnapshotNamespace to ImageCtx Jan 18, 2017

dillaman reviewed Jan 18, 2017

View reviewed changes

VictorDenisov force-pushed the snapshot_namespace_index branch from bcacaf6 to 3b58e62 Compare January 18, 2017 17:46

dillaman reviewed Jan 18, 2017

View reviewed changes

VictorDenisov force-pushed the snapshot_namespace_index branch from baf59c7 to 3f0b60e Compare January 21, 2017 02:13

dillaman reviewed Jan 23, 2017

View reviewed changes

VictorDenisov force-pushed the snapshot_namespace_index branch from 78f614c to 1cb5c9e Compare January 24, 2017 17:45

VictorDenisov force-pushed the snapshot_namespace_index branch from 0c8b865 to 091316d Compare January 27, 2017 00:16

VictorDenisov force-pushed the snapshot_namespace_index branch from ae100ef to d424a64 Compare January 30, 2017 22:17

VictorDenisov changed the title ~~[DNM] librbd: add SnapshotNamespace to ImageCtx~~ librbd: add SnapshotNamespace to ImageCtx Jan 30, 2017

dillaman reviewed Jan 31, 2017

View reviewed changes

VictorDenisov force-pushed the snapshot_namespace_index branch 5 times, most recently from 0dd9226 to e04700f Compare March 6, 2017 09:25

dillaman reviewed Mar 9, 2017

View reviewed changes

VictorDenisov force-pushed the snapshot_namespace_index branch from e04700f to 4a6ca09 Compare March 13, 2017 07:57

VictorDenisov force-pushed the snapshot_namespace_index branch 5 times, most recently from 7b4b5f5 to d6d1747 Compare March 15, 2017 05:27

dillaman reviewed Mar 15, 2017

View reviewed changes

VictorDenisov force-pushed the snapshot_namespace_index branch from d6d1747 to de85e00 Compare March 16, 2017 07:48

dillaman approved these changes Mar 16, 2017

View reviewed changes

dillaman added the wip-jason-testing label Mar 16, 2017

dillaman merged commit de85e00 into ceph:master Mar 17, 2017

dillaman pushed a commit that referenced this pull request Mar 17, 2017

Merge pull request #12970 from VictorDenisov/snapshot_namespace_index

76d46b5

librbd: add SnapshotNamespace to ImageCtx Reviewed-by: Jason Dillaman <dillaman@redhat.com>

VictorDenisov deleted the snapshot_namespace_index branch March 19, 2017 02:03

		@@ -428,11 +430,11 @@ struct C_InvalidateCache : public Context {
		data_ctx.snap_set_read(snap_id);

		@@ -435,11 +437,11 @@ struct C_InvalidateCache : public Context {
		data_ctx.snap_set_read(snap_id);

librbd: add SnapshotNamespace to ImageCtx #12970

librbd: add SnapshotNamespace to ImageCtx #12970

Conversation

VictorDenisov commented Jan 17, 2017 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

VictorDenisov commented Jan 18, 2017

Choose a reason for hiding this comment

dillaman Jan 18, 2017 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

VictorDenisov commented Jan 21, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

dillaman commented Jan 26, 2017

VictorDenisov commented Jan 26, 2017

VictorDenisov commented Jan 27, 2017

VictorDenisov commented Jan 30, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

dillaman commented Jan 31, 2017

VictorDenisov commented Mar 6, 2017

dillaman commented Mar 8, 2017

dillaman Mar 9, 2017 • edited

Choose a reason for hiding this comment

dillaman Mar 9, 2017 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

dillaman Mar 9, 2017 • edited

Choose a reason for hiding this comment

dillaman left a comment

Choose a reason for hiding this comment

dillaman commented Mar 13, 2017

VictorDenisov commented Mar 15, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

VictorDenisov commented Mar 16, 2017

dillaman left a comment

Choose a reason for hiding this comment

dillaman commented Mar 17, 2017

VictorDenisov commented Jan 17, 2017 •

edited

dillaman Jan 18, 2017 •

edited

dillaman Mar 9, 2017 •

edited

dillaman Mar 9, 2017 •

edited

dillaman Mar 9, 2017 •

edited