README: Explain that there are no incremental backups.
[dss.git] / dss.c
diff --git a/dss.c b/dss.c
index 258cd3f..07a6042 100644 (file)
--- a/dss.c
+++ b/dss.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 2008-2011 Andre Noll <maan@systemlinux.org>
+ * Copyright (C) 2008-2011 Andre Noll <maan@tuebingen.mpg.de>
  *
  * Licensed under the GPL v2. For licencing details see COPYING.
  */
@@ -45,6 +45,8 @@ static int signal_pipe;
 static pid_t create_pid;
 /** Whether the pre-create-hook/rsync/post-create-hook is currently stopped. */
 static int create_process_stopped;
+/** How many times in a row the rsync command failed. */
+static int num_consecutive_rsync_errors;
 /** Process id of current pre-remove/rm/post-remove process. */
 static pid_t remove_pid;
 /** When the next snapshot is due. */
@@ -125,6 +127,7 @@ static void dump_dss_config(const char *msg)
                "reference_snapshot: %s\n"
                "snapshot_creation_status: %s\n"
                "snapshot_removal_status: %s\n"
+               "num_consecutive_rsync_errors: %d\n"
                ,
                (int) getpid(),
                logfile? conf.logfile_arg : "stderr",
@@ -135,7 +138,8 @@ static void dump_dss_config(const char *msg)
                name_of_reference_snapshot?
                        name_of_reference_snapshot : "(none)",
                hook_status_description[snapshot_creation_status],
-               hook_status_description[snapshot_removal_status]
+               hook_status_description[snapshot_removal_status],
+               num_consecutive_rsync_errors
        );
        if (create_pid != 0)
                fprintf(log,
@@ -531,27 +535,42 @@ static struct snapshot *find_outdated_snapshot(struct snapshot_list *sl)
 
 static struct snapshot *find_oldest_removable_snapshot(struct snapshot_list *sl)
 {
-       int i;
-       struct snapshot *s;
+       int i, num_complete;
+       struct snapshot *s, *ref = NULL;
+
+       num_complete = num_complete_snapshots(sl);
+       if (num_complete <= conf.min_complete_arg)
+               return NULL;
        FOR_EACH_SNAPSHOT(s, i, sl) {
                if (snapshot_is_being_created(s))
                        continue;
-               if (is_reference_snapshot(s))
+               if (is_reference_snapshot(s)) { /* avoid this one */
+                       ref = s;
                        continue;
+               }
                DSS_INFO_LOG(("oldest removable snapshot: %s\n", s->name));
                return s;
        }
-       return NULL;
+       assert(ref);
+       DSS_WARNING_LOG(("removing reference snapshot %s\n", ref->name));
+       return ref;
 }
 
 static int rename_incomplete_snapshot(int64_t start)
 {
        char *old_name;
        int ret;
+       int64_t now;
 
+       /*
+        * We don't want the dss_rename() below to fail with EEXIST because the
+        * last complete snapshot was created (and completed) in the same
+        * second as this one.
+        */
+       while ((now = get_current_time()) == start)
+               sleep(1);
        free(path_to_last_complete_snapshot);
-       ret = complete_name(start, get_current_time(),
-               &path_to_last_complete_snapshot);
+       ret = complete_name(start, now, &path_to_last_complete_snapshot);
        if (ret < 0)
                return ret;
        old_name = incomplete_name(start);
@@ -829,23 +848,23 @@ static int handle_rsync_exit(int status)
        es = WEXITSTATUS(status);
        /*
         * Restart rsync on non-fatal errors:
-        * 12: Error in rsync protocol data stream
-        * 13: Errors with program diagnostics
+        * 24: Partial transfer due to vanished source files
         */
-       if (es == 12 || es == 13) {
-               DSS_WARNING_LOG(("rsync process %d returned %d -- restarting\n",
-                       (int)create_pid, es));
+       if (es != 0 && es != 24) {
+               DSS_WARNING_LOG(("rsync exit code %d, error count %d\n",
+                       es, ++num_consecutive_rsync_errors));
+               if (num_consecutive_rsync_errors > conf.max_rsync_errors_arg) {
+                       ret = -E_TOO_MANY_RSYNC_ERRORS;
+                       snapshot_creation_status = HS_READY;
+                       goto out;
+               }
+               DSS_WARNING_LOG(("restarting rsync process\n"));
                snapshot_creation_status = HS_NEEDS_RESTART;
                next_snapshot_time = get_current_time() + 60;
                ret = 1;
                goto out;
        }
-       if (es != 0 && es != 23 && es != 24) {
-               DSS_ERROR_LOG(("rsync process %d returned %d\n", (int)create_pid, es));
-               ret = -E_BAD_EXIT_CODE;
-               snapshot_creation_status = HS_READY;
-               goto out;
-       }
+       num_consecutive_rsync_errors = 0;
        ret = rename_incomplete_snapshot(current_snapshot_creation_time);
        if (ret < 0)
                goto out;
@@ -1031,6 +1050,13 @@ static int handle_sighup(void)
        return change_to_dest_dir();
 }
 
+static void kill_children(void)
+{
+       restart_create_process();
+       dss_kill(create_pid, SIGTERM, NULL);
+       dss_kill(remove_pid, SIGTERM, NULL);
+}
+
 static int handle_signal(void)
 {
        int sig, ret = next_signal();
@@ -1041,9 +1067,7 @@ static int handle_signal(void)
        switch (sig) {
        case SIGINT:
        case SIGTERM:
-               restart_create_process();
-               dss_kill(create_pid, SIGTERM, NULL);
-               dss_kill(remove_pid, SIGTERM, NULL);
+               kill_children();
                ret = -E_SIGNAL;
                break;
        case SIGHUP:
@@ -1141,7 +1165,7 @@ static void create_rsync_argv(char ***argv, int64_t *num)
 
        *argv = dss_malloc((15 + conf.rsync_option_given) * sizeof(char *));
        (*argv)[i++] = dss_strdup("rsync");
-       (*argv)[i++] = dss_strdup("-aq");
+       (*argv)[i++] = dss_strdup("-a");
        (*argv)[i++] = dss_strdup("--delete");
        for (j = 0; j < conf.rsync_option_given; j++)
                (*argv)[i++] = dss_strdup(conf.rsync_option_arg[j]);
@@ -1312,6 +1336,7 @@ static int com_run(void)
        ret = select_loop();
        if (ret >= 0) /* impossible */
                ret = -E_BUG;
+       kill_children();
        exit_hook(ret);
        return ret;
 }
@@ -1435,7 +1460,7 @@ static int com_ls(void)
                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;
 }
@@ -1499,6 +1524,7 @@ int main(int argc, char **argv)
        if (ret < 0)
                goto out;
        ret = call_command_handler();
+       signal_shutdown();
 out:
        if (ret < 0)
                DSS_EMERG_LOG(("%s\n", dss_strerror(-ret)));