projects
/
cumulus.git
/ blobdiff
commit
grep
author
committer
pickaxe
?
search:
re
summary
|
shortlog
|
log
|
commit
|
commitdiff
|
tree
raw
|
inline
| side by side
Report compressed size of data written in a backup as well as uncompressed.
[cumulus.git]
/
store.cc
diff --git
a/store.cc
b/store.cc
index
930794d
..
00d0878
100644
(file)
--- a/
store.cc
+++ b/
store.cc
@@
-1,5
+1,5
@@
/* LBS: An LFS-inspired filesystem backup system
/* LBS: An LFS-inspired filesystem backup system
- * Copyright (C) 200
7
Michael Vrable
+ * Copyright (C) 200
8
Michael Vrable
*
* Backup data is stored in a collection of objects, which are grouped together
* into segments for storage purposes. This implementation of the object store
*
* Backup data is stored in a collection of objects, which are grouped together
* into segments for storage purposes. This implementation of the object store
@@
-29,6
+29,7
@@
using std::max;
using std::list;
using std::map;
using std::max;
using std::list;
using std::map;
+using std::pair;
using std::set;
using std::string;
using std::set;
using std::string;
@@
-46,17
+47,15
@@
static void cloexec(int fd)
fcntl(fd, F_SETFD, flags | FD_CLOEXEC);
}
fcntl(fd, F_SETFD, flags | FD_CLOEXEC);
}
-Tarfile::Tarfile(
const string &path
, const string &segment)
+Tarfile::Tarfile(
RemoteFile *file
, const string &segment)
: size(0),
segment_name(segment)
{
assert(sizeof(struct tar_header) == TAR_BLOCK_SIZE);
: size(0),
segment_name(segment)
{
assert(sizeof(struct tar_header) == TAR_BLOCK_SIZE);
- real_fd = open(path.c_str(), O_WRONLY | O_CREAT, 0666);
- if (real_fd < 0)
- throw IOException("Error opening output file");
-
- filter_fd = spawn_filter(real_fd);
+ this->file = file;
+ real_fd = file->get_fd();
+ filter_fd = spawn_filter(real_fd, filter_program, &filter_pid);
}
Tarfile::~Tarfile()
}
Tarfile::~Tarfile()
@@
-86,9
+85,10
@@
Tarfile::~Tarfile()
* on the TAR output. The file descriptor to which output should be written
* must be specified; the return value is the file descriptor which will be
* attached to the standard input of the filter program. */
* on the TAR output. The file descriptor to which output should be written
* must be specified; the return value is the file descriptor which will be
* attached to the standard input of the filter program. */
-int
Tarfile::spawn_filter(int fd_out
)
+int
spawn_filter(int fd_out, const char *program, pid_t *filter_pid
)
{
int fds[2];
{
int fds[2];
+ pid_t pid;
/* Create a pipe for communicating with the filter process. */
if (pipe(fds) < 0) {
/* Create a pipe for communicating with the filter process. */
if (pipe(fds) < 0) {
@@
-96,14
+96,16
@@
int Tarfile::spawn_filter(int fd_out)
}
/* Create a child process which can exec() the filter program. */
}
/* Create a child process which can exec() the filter program. */
-
filter_
pid = fork();
- if (
filter_
pid < 0)
+ pid = fork();
+ if (pid < 0)
throw IOException("Unable to fork filter process");
throw IOException("Unable to fork filter process");
- if (
filter_
pid > 0) {
+ if (pid > 0) {
/* Parent process */
close(fds[0]);
cloexec(fds[1]);
/* Parent process */
close(fds[0]);
cloexec(fds[1]);
+ if (filter_pid != NULL)
+ *filter_pid = pid;
} else {
/* Child process. Rearrange file descriptors. stdin is fds[0], stdout
* is fd_out, stderr is unchanged. */
} else {
/* Child process. Rearrange file descriptors. stdin is fds[0], stdout
* is fd_out, stderr is unchanged. */
@@
-118,7
+120,7
@@
int Tarfile::spawn_filter(int fd_out)
close(fd_out);
/* Exec the filter program. */
close(fd_out);
/* Exec the filter program. */
- execlp("/bin/sh", "/bin/sh", "-c",
filter_
program, NULL);
+ execlp("/bin/sh", "/bin/sh", "-c", program, NULL);
/* Should not reach here except for error cases. */
fprintf(stderr, "Could not exec filter: %m\n");
/* Should not reach here except for error cases. */
fprintf(stderr, "Could not exec filter: %m\n");
@@
-149,19
+151,13
@@
void Tarfile::tar_write(const char *data, size_t len)
void Tarfile::write_object(int id, const char *data, size_t len)
{
void Tarfile::write_object(int id, const char *data, size_t len)
{
+ struct tar_header header;
+ memset(&header, 0, sizeof(header));
+
char buf[64];
sprintf(buf, "%08x", id);
string path = segment_name + "/" + buf;
char buf[64];
sprintf(buf, "%08x", id);
string path = segment_name + "/" + buf;
- internal_write_object(path, data, len);
-}
-
-void Tarfile::internal_write_object(const string &path,
- const char *data, size_t len)
-{
- struct tar_header header;
- memset(&header, 0, sizeof(header));
-
assert(path.size() < 100);
memcpy(header.name, path.data(), path.size());
sprintf(header.mode, "%07o", 0600);
assert(path.size() < 100);
memcpy(header.name, path.data(), path.size());
sprintf(header.mode, "%07o", 0600);
@@
-217,7
+213,8
@@
size_t Tarfile::size_estimate()
static const size_t SEGMENT_SIZE = 4 * 1024 * 1024;
static const size_t SEGMENT_SIZE = 4 * 1024 * 1024;
-static map<string, int64_t> group_sizes;
+/* Backup size summary: segment type -> (uncompressed size, compressed size) */
+static map<string, pair<int64_t, int64_t> > group_sizes;
ObjectReference TarSegmentStore::write_object(const char *data, size_t len,
const std::string &group)
ObjectReference TarSegmentStore::write_object(const char *data, size_t len,
const std::string &group)
@@
-230,11
+227,13
@@
ObjectReference TarSegmentStore::write_object(const char *data, size_t len,
segment = new segment_info;
segment->name = generate_uuid();
segment = new segment_info;
segment->name = generate_uuid();
+ segment->group = group;
segment->basename = segment->name + ".tar";
segment->basename += filter_extension;
segment->basename = segment->name + ".tar";
segment->basename += filter_extension;
- segment->fullname = path + "/" + segment->basename;
- segment->file = new Tarfile(segment->fullname, segment->name);
segment->count = 0;
segment->count = 0;
+ segment->size = 0;
+ segment->rf = remote->alloc_file(segment->basename, "segments");
+ segment->file = new Tarfile(segment->rf, segment->name);
segments[group] = segment;
} else {
segments[group] = segment;
} else {
@@
-247,8
+246,9
@@
ObjectReference TarSegmentStore::write_object(const char *data, size_t len,
segment->file->write_object(id, data, len);
segment->count++;
segment->file->write_object(id, data, len);
segment->count++;
+ segment->size += len;
- group_sizes[group] += len;
+ group_sizes[group]
.first
+= len;
ObjectReference ref(segment->name, id_buf);
ObjectReference ref(segment->name, id_buf);
@@
-269,9
+269,10
@@
void TarSegmentStore::sync()
void TarSegmentStore::dump_stats()
{
printf("Data written:\n");
void TarSegmentStore::dump_stats()
{
printf("Data written:\n");
- for (map<string,
int64_t
>::iterator i = group_sizes.begin();
+ for (map<string,
pair<int64_t, int64_t>
>::iterator i = group_sizes.begin();
i != group_sizes.end(); ++i) {
i != group_sizes.end(); ++i) {
- printf(" %s: %lld\n", i->first.c_str(), i->second);
+ printf(" %s: %lld (%lld compressed)\n", i->first.c_str(),
+ i->second.first, i->second.second);
}
}
}
}
@@
-283,12
+284,20
@@
void TarSegmentStore::close_segment(const string &group)
if (db != NULL) {
SHA1Checksum segment_checksum;
if (db != NULL) {
SHA1Checksum segment_checksum;
- if (segment_checksum.process_file(segment->
fullname
.c_str())) {
+ if (segment_checksum.process_file(segment->
rf->get_local_path()
.c_str())) {
string checksum = segment_checksum.checksum_str();
string checksum = segment_checksum.checksum_str();
- db->SetSegmentChecksum(segment->name, segment->basename, checksum);
+ db->SetSegmentChecksum(segment->name, segment->basename, checksum,
+ segment->size);
+ }
+
+ struct stat stat_buf;
+ if (stat(segment->rf->get_local_path().c_str(), &stat_buf) == 0) {
+ group_sizes[segment->group].second += stat_buf.st_size;
}
}
}
}
+ segment->rf->send();
+
segments.erase(segments.find(group));
delete segment;
}
segments.erase(segments.find(group));
delete segment;
}