mds: change the type of data_pools

	Change the type of data_pools from `set` to `vector`

	In following scenario, we found something strange here.

	Here we have two pools as following:

		- data_pool1    id: 20
		- data_pool2    id: 21

	First, we create ceph fs with `data_pool2` and default data_pool is `data_pool2`.
	Then, add the new data pool `data_pool1`.
	Now, the default data_pool will be data_pool1.

	Because set will sort, we can't remain the default data_pool as we wish.
	(we assume that the default data_pool should be the data_pool that we create cephfs)

	It seems vector is more suitable than set.

Signed-off-by: Vicente Cheng <vicente_cheng@bigtera.com>

1 files changed, 3 insertions, 5 deletions

diff --git a/src/mds/SnapServer.cc b/src/mds/SnapServer.cc
index 4518abef6cf..ea78bff1d08 100644
--- a/src/mds/SnapServer.cc
+++ b/src/mds/SnapServer.cc
@@ -166,11 +166,9 @@ bool SnapServer::_commit(version_t tid, MMDSTableRequest *req)
     dout(7) << "commit " << tid << " destroy " << sn << " seq " << seq << dendl;
     snaps.erase(sn);
 
-    for (set<int64_t>::const_iterator p = mds->mdsmap->get_data_pools().begin();
-	 p != mds->mdsmap->get_data_pools().end();
-	 ++p) {
-      need_to_purge[*p].insert(sn);
-      need_to_purge[*p].insert(seq);
+    for (const auto p : mds->mdsmap->get_data_pools()) {
+      need_to_purge[p].insert(sn);
+      need_to_purge[p].insert(seq);
     }
 
     pending_destroy.erase(tid);