X-Git-Url: http://git.vrable.net/?a=blobdiff_plain;f=scandir.cc;h=65151bca1afbe620b594d3ba4834935639d81617;hb=e29cf5eefb0ff3be77a548edb6bbd89ea58c02a9;hp=d39029cd57197933eed2e8e57ed8a6bc23ca3899;hpb=bae7cf459803abce603c8b3d3009153936378b42;p=cumulus.git diff --git a/scandir.cc b/scandir.cc index d39029c..65151bc 100644 --- a/scandir.cc +++ b/scandir.cc @@ -57,6 +57,11 @@ LocalDb *db; /* Keep track of all segments which are needed to reconstruct the snapshot. */ std::set segment_list; +/* Snapshot intent: 1=daily, 7=weekly, etc. This is not used directly, but is + * stored in the local database and can help guide segment cleaning and + * snapshot expiration policies. */ +double snapshot_intent = 1.0; + /* Selection of files to include/exclude in the snapshot. */ std::list includes; // Paths in which files should be saved std::list excludes; // Paths which will not be saved @@ -114,7 +119,7 @@ int64_t dumpfile(int fd, dictionary &file_info, const string &path, * re-reading the entire contents. */ bool cached = false; - if (metawriter->matched() && metawriter->is_unchanged(&stat_buf)) { + if (metawriter->find(path) && metawriter->is_unchanged(&stat_buf)) { cached = true; list blocks = metawriter->get_blocks(); @@ -137,7 +142,8 @@ int64_t dumpfile(int fd, dictionary &file_info, const string &path, i != blocks.end(); ++i) { const ObjectReference &ref = *i; object_list.push_back(ref.to_string()); - segment_list.insert(ref.get_segment()); + if (ref.is_normal()) + add_segment(ref.get_segment()); db->UseObject(ref); } size = stat_buf.st_size; @@ -160,16 +166,34 @@ int64_t dumpfile(int fd, dictionary &file_info, const string &path, hash.process(block_buf, bytes); + // Sparse file processing: if we read a block of all zeroes, encode + // that explicitly. + bool all_zero = true; + for (int i = 0; i < bytes; i++) { + if (block_buf[i] != 0) { + all_zero = false; + break; + } + } + // Either find a copy of this block in an already-existing segment, // or index it so it can be re-used in the future double block_age = 0.0; + ObjectReference ref; + SHA1Checksum block_hash; block_hash.process(block_buf, bytes); string block_csum = block_hash.checksum_str(); - ObjectReference ref = db->FindObject(block_csum, bytes); + + if (all_zero) { + ref = ObjectReference(ObjectReference::REF_ZERO); + ref.set_range(0, bytes); + } else { + ref = db->FindObject(block_csum, bytes); + } // Store a copy of the object if one does not yet exist - if (ref.get_segment().size() == 0) { + if (ref.is_null()) { LbsObject *o = new LbsObject; int object_group; @@ -204,11 +228,13 @@ int64_t dumpfile(int fd, dictionary &file_info, const string &path, o->write(tss); ref = o->get_ref(); db->StoreObject(ref, block_csum, bytes, block_age); + ref.set_range(0, bytes); delete o; } object_list.push_back(ref.to_string()); - segment_list.insert(ref.get_segment()); + if (ref.is_normal()) + add_segment(ref.get_segment()); db->UseObject(ref); size += bytes; @@ -248,16 +274,20 @@ void dump_inode(const string& path, // Path within snapshot ssize_t len; printf("%s\n", path.c_str()); - metawriter->find(path); - file_info["path"] = uri_encode(path); + file_info["name"] = uri_encode(path); file_info["mode"] = encode_int(stat_buf.st_mode & 07777, 8); file_info["ctime"] = encode_int(stat_buf.st_ctime); file_info["mtime"] = encode_int(stat_buf.st_mtime); file_info["user"] = encode_int(stat_buf.st_uid); file_info["group"] = encode_int(stat_buf.st_gid); + time_t now = time(NULL); + if (now - stat_buf.st_ctime < 30 || now - stat_buf.st_mtime < 30) + if ((stat_buf.st_mode & S_IFMT) != S_IFDIR) + file_info["volatile"] = "1"; + struct passwd *pwd = getpwuid(stat_buf.st_uid); if (pwd != NULL) { file_info["user"] += " (" + uri_encode(pwd->pw_name) + ")"; @@ -322,6 +352,7 @@ void dump_inode(const string& path, // Path within snapshot if (file_size != stat_buf.st_size) { fprintf(stderr, "Warning: Size of %s changed during reading\n", path.c_str()); + file_info["volatile"] = "1"; } break; @@ -532,7 +563,10 @@ void usage(const char *program) " (defaults to \".bz2\")\n" " --signature-filter=COMMAND\n" " program though which to filter descriptor\n" - " --scheme=NAME optional name for this snapshot\n", + " --scheme=NAME optional name for this snapshot\n" + " --intent=FLOAT intended backup type: 1=daily, 7=weekly, ...\n" + " (defaults to \"1\")\n" + " --full-metadata do not re-use metadata from previous backups\n", lbs_version, program ); } @@ -553,6 +587,8 @@ int main(int argc, char *argv[]) {"dest", 1, 0, 0}, // 4 {"scheme", 1, 0, 0}, // 5 {"signature-filter", 1, 0, 0}, // 6 + {"intent", 1, 0, 0}, // 7 + {"full-metadata", 0, 0, 0}, // 8 {NULL, 0, 0, 0}, }; @@ -588,6 +624,14 @@ int main(int argc, char *argv[]) case 6: // --signature-filter signature_filter = optarg; break; + case 7: // --intent + snapshot_intent = atof(optarg); + if (snapshot_intent <= 0) + snapshot_intent = 1; + break; + case 8: // --full-metadata + flag_full_metadata = true; + break; default: fprintf(stderr, "Unhandled long option!\n"); return 1; @@ -658,7 +702,8 @@ int main(int argc, char *argv[]) string database_path = localdb_dir + "/localdb.sqlite"; db = new LocalDb; db->Open(database_path.c_str(), desc_buf, - backup_scheme.size() ? backup_scheme.c_str() : NULL); + backup_scheme.size() ? backup_scheme.c_str() : NULL, + snapshot_intent); tss = new TarSegmentStore(backup_dest, db); @@ -745,6 +790,7 @@ int main(int argc, char *argv[]) fprintf(descriptor, "Date: %s\n", desc_buf); if (backup_scheme.size() > 0) fprintf(descriptor, "Scheme: %s\n", backup_scheme.c_str()); + fprintf(descriptor, "Backup-Intent: %g\n", snapshot_intent); fprintf(descriptor, "Root: %s\n", backup_root.c_str()); SHA1Checksum checksum_csum;