curvefs: support curvebs as data backend

WORKSPACE

@@ -20,7 +20,7 @@ load("@bazel_tools//tools/build_defs/repo:git.bzl", "git_repository")
 
 git_repository(
     name = "com_github_baidu_braft",
-    remote = "https://github.com/baidu/braft",
+    remote = "https://gitee.com/baidu/braft",
     commit = "e255c0e4b18d1a8a5d484d4b647f41ff1385ef1e",
 )
 
@@ -36,7 +36,7 @@ http_archive(
     name = "com_google_protobuf",
     sha256 = "cef7f1b5a7c5fba672bec2a319246e8feba471f04dcebfe362d55930ee7c1c30",
     strip_prefix = "protobuf-3.5.0",
-    urls = ["https://github.com/google/protobuf/archive/v3.5.0.zip"],
+    urls = ["https://curve-build.nos-eastchina1.126.net/protobuf-3.5.0.zip"],
 )
 
 bind(
@@ -48,7 +48,7 @@ bind(
 new_git_repository(
     name = "com_google_googletest",
     build_file = "bazel/gmock.BUILD",
-    remote = "https://github.com/google/googletest",
+    remote = "https://gitee.com/mirrors/googletest",
     tag = "release-1.8.0",
 )
 
@@ -61,7 +61,7 @@ bind(
 # brpc内BUILD文件在依赖glog时, 直接指定的依赖是"@com_github_google_glog//:glog"
 git_repository(
     name = "com_github_google_glog",
-    remote = "https://github.com/google/glog",
+    remote = "https://gitee.com/mirrors/glog",
     commit = "4cc89c9e2b452db579397887c37f302fb28f6ca1",
     patch_args = ["-p1"],
     patches = ["//:thirdparties/glog/glog.patch"],
@@ -77,7 +77,7 @@ http_archive(
     name = "com_github_gflags_gflags",
     strip_prefix = "gflags-2.2.2",
     urls = [
-        "https://mirror.bazel.build/github.com/gflags/gflags/archive/v2.2.2.tar.gz",
+        "https://curve-build.nos-eastchina1.126.net/gflags-2.2.2.tar.gz",
         "https://github.com/gflags/gflags/archive/v2.2.2.tar.gz",
     ],
 )
@@ -91,7 +91,7 @@ new_http_archive(
     name = "com_github_google_leveldb",
     build_file = "bazel/leveldb.BUILD",
     strip_prefix = "leveldb-a53934a3ae1244679f812d998a4f16f2c7f309a6",
-    url = "https://github.com/google/leveldb/archive/a53934a3ae1244679f812d998a4f16f2c7f309a6.tar.gz",
+    url = "https://curve-build.nos-eastchina1.126.net/leveldb-a53934a3ae1244679f812d998a4f16f2c7f309a6.tar.gz",
 )
 
 bind(
@@ -101,7 +101,7 @@ bind(
 
 git_repository(
     name = "com_github_apache_brpc",
-    remote = "https://github.com/apache/incubator-brpc",
+    remote = "https://gitee.com/baidu/BRPC",
     commit = "1b9e00641cbec1c8803da6a1f7f555398c954cb0",
     patches = ["//:thirdparties/brpc/brpc.patch"],
     patch_args = ["-p1"],
@@ -131,7 +131,7 @@ bind(
 new_git_repository(
     name = "jsoncpp",
     build_file = "bazel/jsoncpp.BUILD",
-    remote = "https://github.com/open-source-parsers/jsoncpp.git",
+    remote = "https://gitee.com/mirrors/jsoncpp",
     tag = "1.8.4",
 )
 
@@ -149,7 +149,7 @@ new_local_repository(
 new_http_archive(
     name = "aws",
     urls = [
-        "https://github.com/aws/aws-sdk-cpp/archive/1.7.340.tar.gz",
+        "https://curve-build.nos-eastchina1.126.net/aws-sdk-cpp-1.7.340.tar.gz",
         "https://mirror.bazel.build/github.com/aws/aws-sdk-cpp/archive/1.7.340.tar.gz",
     ],
     sha256 = "2e82517045efb55409cff1408c12829d9e8aea22c1e2888529cb769b7473b0bf",
@@ -160,6 +160,7 @@ new_http_archive(
 new_http_archive(
     name = "aws_c_common",
     urls = [
+        "https://curve-build.nos-eastchina1.126.net/aws-c-common-0.4.29.tar.gz",
         "https://github.com/awslabs/aws-c-common/archive/v0.4.29.tar.gz",
         "https://mirror.tensorflow.org/github.com/awslabs/aws-c-common/archive/v0.4.29.tar.gz",
     ],
@@ -171,6 +172,7 @@ new_http_archive(
 new_http_archive(
     name = "aws_c_event_stream",
     urls = [
+        "https://curve-build.nos-eastchina1.126.net/aws-c-event-stream-0.1.4.tar.gz",
         "https://github.com/awslabs/aws-c-event-stream/archive/v0.1.4.tar.gz",
         "https://mirror.tensorflow.org/github.com/awslabs/aws-c-event-stream/archive/v0.1.4.tar.gz",
     ],
@@ -182,6 +184,7 @@ new_http_archive(
 new_http_archive(
     name = "aws_checksums",
     urls = [
+        "https://curve-build.nos-eastchina1.126.net/aws-checksums-0.1.5.tar.gz",
         "https://github.com/awslabs/aws-checksums/archive/v0.1.5.tar.gz",
         "https://mirror.tensorflow.org/github.com/awslabs/aws-checksums/archive/v0.1.5.tar.gz",
     ],

curvefs/conf/client.conf

@@ -50,7 +50,13 @@ spaceserver.spaceaddr=127.0.0.1:19999  # __ANSIBLE_TEMPLATE__ {{ groups.space |
 spaceserver.rpcTimeoutMs=1000
 
 #### bdev
+# curve client's config file
 bdev.confpath=/etc/curve/client.conf
+# currently, we only use synchronous read/write interface of curve client
+# so, use a dedicated threadpool to issue sync io requests to curve client to
+# avoid performance degradation if single fs IO reqeust forms more than one requests to curve client
+# TODO(wuhanqing): remove this threadpool and use asynchronous interface
+bdev.threadnum=10
 
 #### extentManager
 extentManager.preAllocSize=65536
@@ -79,6 +85,19 @@ volume.bigFileSize=1048576
 volume.volBlockSize=4096
 volume.fsBlockSize=4096
 
+# alloator type, supported {bitmap}
+volume.allocator.type=bitmap
+
+## for bitmap allocator
+# size of each bit, default is 4MiB
+volume.bitmapallocator.size_per_bit=4194304
+
+# small allocation proportion [0-1]
+volume.bitmapallocator.small_alloc_proportion=0.2
+
+# number of block groups that allocated once
+volume.blockgroup.allocate_once=4
+
 #### s3
 # the max size that fuse send
 s3.fuseMaxSize=131072

curvefs/conf/tools.conf

@@ -14,11 +14,14 @@ etcdAddr=127.0.0.1:12379  # __CURVEADM_TEMPLATE__ ${cluster_etcd_addr} __CURVEAD
 blockSize=1048576
 fsType=s3
 # volume
-volumeSize=1048576
+volumeSize=0
 volumeBlockSize=4096
 volumeName=volume
 volumeUser=user
 volumePassword=password
+volumeBlockGroupSize=134217728
+# support |AtStart| and |AtEnd|
+volumeBitmapLocation=AtStart
 # s3
 s3.ak=ak
 s3.sk=sk

curvefs/proto/common.proto

@@ -23,12 +23,22 @@ enum FSType {
     TYPE_S3 = 2;
 }
 
+// bitmap location for each block group
+enum BitmapLocation {
+    AtStart = 1;
+    AtEnd = 2;
+}
+
 message Volume {
     required uint64 volumeSize = 1;
     required uint64 blockSize = 2;
     required string volumeName = 3;
     required string user = 4;
     optional string password = 5;
+    required uint64 blockGroupSize = 6;
+    required BitmapLocation bitmapLocation = 7;
+
+    // TODO(all): maybe we need curvebs cluster's mds ip and port in here
 }
 
 message S3Info {

curvefs/proto/metaserver.proto

@@ -146,10 +146,11 @@ message GetInodeRequest {
 }
 
 enum FsFileType {
-    TYPE_DIRECTORY = 1;
-    TYPE_FILE = 2;
-    TYPE_SYM_LINK = 3;
-    TYPE_S3 = 4;
+    TYPE_DIRECTORY = 1;   // 1 << 0
+    TYPE_FILE = 2;        // 1 << 1
+    TYPE_SYM_LINK = 4;    // 1 << 2
+    TYPE_S3 = 18;         // (1 << 4) | TYPE_FILE
+    TYPE_VOLUME = 34;     // (1 << 5) | TYPE_FILE
 };
 
 message VolumeExtent {
@@ -194,7 +195,7 @@ message Inode {
     required FsFileType type = 14;
     optional string symlink = 15;   // TYPE_SYM_LINK only
     optional uint64 rdev = 16;
-    optional VolumeExtentList volumeExtentList = 17;  // TYPE_FILE only
+    map<uint64, VolumeExtentList> volumeExtentMap = 17;  // TYPE_FILE only, key is file's offset
     map<uint64, S3ChunkInfoList> s3ChunkInfoMap = 18; // TYPE_S3 only, first is chunk index
     optional uint32 dtime = 19;
     optional uint32 openmpcount = 20; // openmpcount mount points had the file open
@@ -266,7 +267,7 @@ message UpdateInodeRequest {
     optional uint32 uid = 13;
     optional uint32 gid = 14;
     optional uint32 mode = 15;
-    optional VolumeExtentList volumeExtentList = 16;
+    map<uint64, VolumeExtentList> volumeExtentMap = 16;
     map<uint64, S3ChunkInfoList> s3ChunkInfoMap = 17;
     optional uint32 nlink = 18;
     optional InodeOpenStatusChange inodeOpenstatusChange = 19;