static FILE *logfile;
/** The read end of the signal pipe */
static int signal_pipe;
-/** Process id of current rsync process. */
-static pid_t rsync_pid;
-/** Whether the rsync process is currently stopped */
-static int rsync_stopped;
-/** Process id of current rm process. */
-static pid_t rm_pid;
+/** Process id of current pre-create-hook/rsync/post-create-hook process. */
+static pid_t create_pid;
+/** Whether the pre-create-hook/rsync/post-create-hook is currently stopped. */
+static int create_process_stopped;
+/** Process id of current pre-remove/rm/post-remove process. */
+static pid_t remove_pid;
/** When the next snapshot is due. */
static struct timeval next_snapshot_time;
-/** The pid of the pre-create hook. */
-static pid_t pre_create_hook_pid;
-/** The pid of the post-create hook. */
-static pid_t post_create_hook_pid;
/** Creation time of the snapshot currently being created. */
static int64_t current_snapshot_creation_time;
/** Needed by the post-create hook. */
static char *path_to_last_complete_snapshot;
/** \sa \ref snap.h for details. */
static unsigned snapshot_creation_status;
+/** \sa \ref snap.h for details. */
+static unsigned snapshot_removal_status;
DEFINE_DSS_ERRLIST;
tmp.tv_sec = x;
tmp.tv_usec = 0;
ret = tv_diff(&unit_interval, &tmp, &diff); /* total sleep time per unit interval */
- gettimeofday(&now, NULL);
- if (ret < 0 || !s)
+ if (ret < 0 || !s) /* unit_interval < tmp or no snapshot */
goto min_sleep;
tv_divide(wanted, &diff, &tmp); /* sleep time betweeen two snapshots */
- diff.tv_sec = s->completion_time;
+ diff.tv_sec = s->completion_time; /* completion time of the the latest snaphot */
diff.tv_usec = 0;
tv_add(&diff, &tmp, &next_snapshot_time);
+ gettimeofday(&now, NULL);
if (tv_diff(&now, &next_snapshot_time, NULL) < 0)
goto out;
min_sleep:
}
-static int remove_snapshot(struct snapshot *s)
+static int remove_snapshot(struct snapshot *s, char *why)
{
int fds[3] = {0, 0, 0};
- assert(!rm_pid);
+ assert(!remove_pid);
char *new_name = being_deleted_name(s);
int ret = dss_rename(s->name, new_name);
char *argv[] = {"rm", "-rf", new_name, NULL};
if (ret < 0)
goto out;
- DSS_NOTICE_LOG("removing %s (interval = %i)\n", s->name, s->interval);
- ret = dss_exec(&rm_pid, argv[0], argv, fds);
+ DSS_NOTICE_LOG("removing %s snapshot %s (interval = %i)\n",
+ why, s->name, s->interval);
+ ret = dss_exec(&remove_pid, argv[0], argv, fds);
out:
free(new_name);
return ret;
}
+static int snapshot_is_being_created(struct snapshot *s)
+{
+ return s->creation_time == current_snapshot_creation_time;
+}
+
/*
* return: 0: no redundant snapshots, 1: rm process started, negative: error
*/
FOR_EACH_SNAPSHOT(s, i, sl) {
int64_t this_score;
+ if (snapshot_is_being_created(s))
+ continue;
//DSS_DEBUG_LOG("checking %s\n", s->name);
if (s->interval > interval) {
prev = s;
victim->name, victim->interval);
continue;
}
- ret = remove_snapshot(victim);
+ ret = remove_snapshot(victim, "redundant");
return ret < 0? ret : 1;
}
return 0;
DSS_DEBUG_LOG("looking for snapshots belonging to intervals greater than %d\n",
conf.num_intervals_arg);
FOR_EACH_SNAPSHOT(s, i, sl) {
+ if (snapshot_is_being_created(s))
+ continue;
if (s->interval <= conf.num_intervals_arg)
continue;
if (conf.dry_run_given) {
s->name, s->interval);
continue;
}
- ret = remove_snapshot(s);
+ ret = remove_snapshot(s, "outdated");
if (ret < 0)
return ret;
return 1;
if (!s) /* no snapshot found */
return 0;
DSS_INFO_LOG("oldest snapshot: %s\n", s->name);
- if (s->creation_time == current_snapshot_creation_time)
- return 0; /* do not remove the snapshot currently being created */
- return remove_snapshot(s);
+ if (snapshot_is_being_created(s))
+ return 0;
+ return remove_snapshot(s, "oldest");
}
static int rename_incomplete_snapshot(int64_t start)
snapshot_creation_status = SCS_PRE_HOOK_SUCCESS;
return 0;
}
- DSS_NOTICE_LOG("executing %s\n", conf.pre_create_hook_arg);
- ret = dss_exec_cmdline_pid(&pre_create_hook_pid,
+ DSS_DEBUG_LOG("executing %s\n", conf.pre_create_hook_arg);
+ ret = dss_exec_cmdline_pid(&create_pid,
conf.pre_create_hook_arg, fds);
if (ret < 0)
return ret;
compute_next_snapshot_time();
return 0;
}
- cmd = make_message("%s %s", conf.post_create_hook_arg,
- path_to_last_complete_snapshot);
+ cmd = make_message("%s %s/%s", conf.post_create_hook_arg,
+ conf.dest_dir_arg, path_to_last_complete_snapshot);
DSS_NOTICE_LOG("executing %s\n", cmd);
- ret = dss_exec_cmdline_pid(&post_create_hook_pid, cmd, fds);
+ ret = dss_exec_cmdline_pid(&create_pid, cmd, fds);
free(cmd);
if (ret < 0)
return ret;
kill(pid, SIGTERM);
}
-static void stop_rsync_process(void)
+static void stop_create_process(void)
{
- if (!rsync_pid || rsync_stopped)
+ if (!create_pid || create_process_stopped)
return;
- kill(SIGSTOP, rsync_pid);
- rsync_stopped = 1;
+ kill(SIGSTOP, create_pid);
+ create_process_stopped = 1;
}
-static void restart_rsync_process(void)
+static void restart_create_process(void)
{
- if (!rsync_pid || !rsync_stopped)
+ if (!create_pid || !create_process_stopped)
return;
- kill (SIGCONT, rsync_pid);
- rsync_stopped = 0;
+ kill (SIGCONT, create_pid);
+ create_process_stopped = 0;
}
-
/**
* Print a log message about the exit status of a child.
*/
static int handle_rm_exit(int status)
{
- rm_pid = 0;
+ remove_pid = 0;
if (!WIFEXITED(status))
return -E_INVOLUNTARY_EXIT;
if (WEXITSTATUS(status))
static int wait_for_rm_process(void)
{
- int status, ret = wait_for_process(rm_pid, &status);
+ int status, ret = wait_for_process(remove_pid, &status);
if (ret < 0)
return ret;
int es, ret;
if (!WIFEXITED(status)) {
- DSS_ERROR_LOG("rsync process %d died involuntary\n", (int)rsync_pid);
+ DSS_ERROR_LOG("rsync process %d died involuntary\n", (int)create_pid);
ret = -E_INVOLUNTARY_EXIT;
snapshot_creation_status = SCS_READY;
compute_next_snapshot_time();
goto out;
}
es = WEXITSTATUS(status);
+ if (es == 13) { /* Errors with program diagnostics */
+ DSS_WARNING_LOG("rsync process %d returned %d -- restarting\n",
+ (int)create_pid, es);
+ snapshot_creation_status = SCS_RSYNC_NEEDS_RESTART;
+ gettimeofday(&next_snapshot_time, NULL);
+ next_snapshot_time.tv_sec += 60;
+ ret = 1;
+ goto out;
+ }
if (es != 0 && es != 23 && es != 24) {
- DSS_ERROR_LOG("rsync process %d returned %d\n", (int)rsync_pid, es);
+ DSS_ERROR_LOG("rsync process %d returned %d\n", (int)create_pid, es);
ret = -E_BAD_EXIT_CODE;
snapshot_creation_status = SCS_READY;
compute_next_snapshot_time();
goto out;
snapshot_creation_status = SCS_RSYNC_SUCCESS;
out:
- rsync_pid = 0;
- rsync_stopped = 0;
+ create_pid = 0;
+ create_process_stopped = 0;
return ret;
}
static int handle_pre_create_hook_exit(int status)
{
int es, ret;
+ static int warn_count;
if (!WIFEXITED(status)) {
snapshot_creation_status = SCS_READY;
}
es = WEXITSTATUS(status);
if (es) {
+ if (!warn_count--) {
+ DSS_NOTICE_LOG("pre_create_hook %s returned %d\n",
+ conf.pre_create_hook_arg, es);
+ DSS_NOTICE_LOG("deferring snapshot creation...\n");
+ warn_count = 60; /* warn only once per hour */
+ }
snapshot_creation_status = SCS_READY;
compute_next_snapshot_time();
- ret = -E_BAD_EXIT_CODE;
+ ret = 0;
goto out;
}
+ warn_count = 0;
snapshot_creation_status = SCS_PRE_HOOK_SUCCESS;
ret = 1;
out:
- pre_create_hook_pid = 0;
+ create_pid = 0;
return ret;
}
if (ret <= 0)
return ret;
- if (pid == rsync_pid)
- return handle_rsync_exit(status);
- if (pid == rm_pid)
- return handle_rm_exit(status);
- if (pid == pre_create_hook_pid)
- return handle_pre_create_hook_exit(status);
- if (pid == post_create_hook_pid) {
- snapshot_creation_status = SCS_READY;
- compute_next_snapshot_time();
- return 1;
+
+ if (pid == create_pid) {
+ switch (snapshot_creation_status) {
+ case SCS_PRE_HOOK_RUNNING:
+ return handle_pre_create_hook_exit(status);
+ case SCS_RSYNC_RUNNING:
+ return handle_rsync_exit(status);
+ case SCS_POST_HOOK_RUNNING:
+ snapshot_creation_status = SCS_READY;
+ compute_next_snapshot_time();
+ return 1;
+ default:
+ DSS_EMERG_LOG("BUG: create can't die in status %d\n",
+ snapshot_creation_status);
+ return -E_BUG;
+ }
}
+ if (pid == remove_pid)
+ return handle_rm_exit(status);
DSS_EMERG_LOG("BUG: unknown process %d died\n", (int)pid);
return -E_BUG;
}
.check_ambiguity = 0,
.print_errors = 1
};
+ if (override) { /* invalidate all rsync options */
+ int i;
+
+ for (i = 0; i < conf.rsync_option_given; i++) {
+ free(conf.rsync_option_arg[i]);
+ conf.rsync_option_arg[i] = NULL;
+ }
+ conf.rsync_option_given = 0;
+ }
cmdline_parser_config_file(config_file, &conf, ¶ms);
}
ret = check_config();
switch (sig) {
case SIGINT:
case SIGTERM:
- restart_rsync_process();
- kill_process(rsync_pid);
- kill_process(rm_pid);
+ restart_create_process();
+ kill_process(create_pid);
+ kill_process(remove_pid);
ret = -E_SIGNAL;
break;
case SIGHUP:
return ret;
}
+/*
+ * We can not use rsync locally if the local user is different from the remote
+ * user or if the src dir is not on the local host (or both).
+ */
+static int use_rsync_locally(char *logname)
+{
+ char *h = conf.remote_host_arg;
+
+ if (strcmp(h, "localhost") && strcmp(h, "127.0.0.1"))
+ return 0;
+ if (conf.remote_user_given && strcmp(conf.remote_user_arg, logname))
+ return 0;
+ return 1;
+}
+
static void create_rsync_argv(char ***argv, int64_t *num)
{
char *logname, *newest;
} else
DSS_INFO_LOG("no previous snapshot found\n");
logname = dss_logname();
- if (conf.remote_user_given && !strcmp(conf.remote_user_arg, logname))
+ if (use_rsync_locally(logname))
(*argv)[i++] = dss_strdup(conf.source_dir_arg);
else
(*argv)[i++] = make_message("%s@%s:%s/", conf.remote_user_given?
static void free_rsync_argv(char **argv)
{
int i;
+
+ if (!argv)
+ return;
for (i = 0; argv[i]; i++)
free(argv[i]);
free(argv);
name = incomplete_name(current_snapshot_creation_time);
DSS_NOTICE_LOG("creating new snapshot %s\n", name);
free(name);
- ret = dss_exec(&rsync_pid, argv[0], argv, fds);
+ ret = dss_exec(&create_pid, argv[0], argv, fds);
if (ret < 0)
return ret;
snapshot_creation_status = SCS_RSYNC_RUNNING;
int ret;
/* check every 60 seconds for free disk space */
struct timeval tv;
+ char **rsync_argv = NULL;
for (;;) {
fd_set rfds;
int low_disk_space;
- char **rsync_argv;
struct timeval now, *tvp;
- if (rm_pid)
+ if (remove_pid)
tvp = NULL; /* sleep until rm process dies */
else { /* sleep one minute */
tv.tv_sec = 60;
if (ret < 0)
goto out;
}
- if (rm_pid)
+ if (remove_pid)
continue;
ret = disk_space_low();
if (ret < 0)
ret = try_to_free_disk_space(low_disk_space);
if (ret < 0)
goto out;
- if (rm_pid) {
- stop_rsync_process();
+ if (remove_pid) {
+ stop_create_process();
continue;
}
- restart_rsync_process();
+ restart_create_process();
gettimeofday(&now, NULL);
if (tv_diff(&next_snapshot_time, &now, NULL) > 0)
continue;
continue;
case SCS_PRE_HOOK_RUNNING:
continue;
+ case SCS_RSYNC_NEEDS_RESTART:
+ ret = create_snapshot(rsync_argv);
+ if (ret < 0)
+ goto out;
+ continue;
case SCS_PRE_HOOK_SUCCESS:
+ free_rsync_argv(rsync_argv);
create_rsync_argv(&rsync_argv, ¤t_snapshot_creation_time);
ret = create_snapshot(rsync_argv);
- free_rsync_argv(rsync_argv);
if (ret < 0)
goto out;
continue;
ret = pre_create_hook();
if (ret < 0)
return ret;
- if (pre_create_hook_pid) {
- ret = wait_for_process(pre_create_hook_pid, &status);
+ if (create_pid) {
+ ret = wait_for_process(create_pid, &status);
if (ret < 0)
return ret;
ret = handle_pre_create_hook_exit(status);
- if (ret < 0)
+ if (ret <= 0) /* error, or pre-create failed */
return ret;
}
create_rsync_argv(&rsync_argv, ¤t_snapshot_creation_time);
ret = create_snapshot(rsync_argv);
if (ret < 0)
goto out;
- ret = wait_for_process(rsync_pid, &status);
+ ret = wait_for_process(create_pid, &status);
if (ret < 0)
goto out;
ret = handle_rsync_exit(status);
if (ret < 0)
goto out;
post_create_hook();
- if (post_create_hook_pid)
- ret = wait_for_process(post_create_hook_pid, &status);
+ if (create_pid)
+ ret = wait_for_process(create_pid, &status);
out:
free_rsync_argv(rsync_argv);
return ret;
struct snapshot *s;
dss_get_snapshot_list(&sl);
- FOR_EACH_SNAPSHOT(s, i, &sl)
- dss_msg("%u\t%s\n", s->interval, s->name);
+ FOR_EACH_SNAPSHOT(s, i, &sl) {
+ int64_t d = 0;
+ if (s->flags & SS_COMPLETE)
+ d = (s->completion_time - s->creation_time) / 60;
+ dss_msg("%u\t%s\t%3" PRId64 ":%02" PRId64 "\n", s->interval, s->name, d/60, d%60);
+ };
free_snapshot_list(&sl);
return 1;
}