diff --git a/cfg/tarantool_silverbox_cfg.c b/cfg/tarantool_silverbox_cfg.c
index 361478850f5fde102e49a027ed5fd257098ed4bc..10c9d296c1856accf346a4787e31a655a31e7dee 100644
--- a/cfg/tarantool_silverbox_cfg.c
+++ b/cfg/tarantool_silverbox_cfg.c
@@ -60,6 +60,8 @@ fill_default_tarantool_cfg(tarantool_cfg *c) {
c->wal_writer_inbox_size = 128;
c->local_hot_standby = 0;
c->wal_dir_rescan_delay = 0.1;
+ c->panic_on_snap_error = 1;
+ c->panic_on_wal_error = 0;
c->remote_hot_standby = 0;
c->wal_feeder_ipaddr = NULL;
c->wal_feeder_port = 0;
@@ -183,6 +185,12 @@ static NameAtom _name__local_hot_standby[] = {
static NameAtom _name__wal_dir_rescan_delay[] = {
{ "wal_dir_rescan_delay", -1, NULL }
};
+static NameAtom _name__panic_on_snap_error[] = {
+ { "panic_on_snap_error", -1, NULL }
+};
+static NameAtom _name__panic_on_wal_error[] = {
+ { "panic_on_wal_error", -1, NULL }
+};
static NameAtom _name__remote_hot_standby[] = {
{ "remote_hot_standby", -1, NULL }
};
@@ -547,6 +555,28 @@ acceptValue(tarantool_cfg* c, OptDef* opt, int check_rdonly) {
if ( (c->wal_dir_rescan_delay == 0 || c->wal_dir_rescan_delay == -HUGE_VAL || c->wal_dir_rescan_delay == HUGE_VAL) && errno == ERANGE)
return CNF_WRONGRANGE;
}
+ else if ( cmpNameAtoms( opt->name, _name__panic_on_snap_error) ) {
+ if (opt->paramType != numberType )
+ return CNF_WRONGTYPE;
+ errno = 0;
+ long int i32 = strtol(opt->paramValue.numberval, NULL, 10);
+ if (i32 == 0 && errno == EINVAL)
+ return CNF_WRONGINT;
+ if ( (i32 == LONG_MIN || i32 == LONG_MAX) && errno == ERANGE)
+ return CNF_WRONGRANGE;
+ c->panic_on_snap_error = i32;
+ }
+ else if ( cmpNameAtoms( opt->name, _name__panic_on_wal_error) ) {
+ if (opt->paramType != numberType )
+ return CNF_WRONGTYPE;
+ errno = 0;
+ long int i32 = strtol(opt->paramValue.numberval, NULL, 10);
+ if (i32 == 0 && errno == EINVAL)
+ return CNF_WRONGINT;
+ if ( (i32 == LONG_MIN || i32 == LONG_MAX) && errno == ERANGE)
+ return CNF_WRONGRANGE;
+ c->panic_on_wal_error = i32;
+ }
else if ( cmpNameAtoms( opt->name, _name__remote_hot_standby) ) {
if (opt->paramType != numberType )
return CNF_WRONGTYPE;
@@ -787,6 +817,8 @@ typedef enum IteratorState {
S_name__wal_writer_inbox_size,
S_name__local_hot_standby,
S_name__wal_dir_rescan_delay,
+ S_name__panic_on_snap_error,
+ S_name__panic_on_wal_error,
S_name__remote_hot_standby,
S_name__wal_feeder_ipaddr,
S_name__wal_feeder_port,
@@ -1148,6 +1180,28 @@ tarantool_cfg_iterator_next(tarantool_cfg_iterator_t* i, tarantool_cfg *c, char
}
sprintf(*v, "%g", c->wal_dir_rescan_delay);
snprintf(buf, PRINTBUFLEN-1, "wal_dir_rescan_delay");
+ i->state = S_name__panic_on_snap_error;
+ return buf;
+ case S_name__panic_on_snap_error:
+ *v = malloc(32);
+ if (*v == NULL) {
+ free(i);
+ out_warning(CNF_NOMEMORY, "No memory to output value");
+ return NULL;
+ }
+ sprintf(*v, "%"PRId32, c->panic_on_snap_error);
+ snprintf(buf, PRINTBUFLEN-1, "panic_on_snap_error");
+ i->state = S_name__panic_on_wal_error;
+ return buf;
+ case S_name__panic_on_wal_error:
+ *v = malloc(32);
+ if (*v == NULL) {
+ free(i);
+ out_warning(CNF_NOMEMORY, "No memory to output value");
+ return NULL;
+ }
+ sprintf(*v, "%"PRId32, c->panic_on_wal_error);
+ snprintf(buf, PRINTBUFLEN-1, "panic_on_wal_error");
i->state = S_name__remote_hot_standby;
return buf;
case S_name__remote_hot_standby:
diff --git a/cfg/tarantool_silverbox_cfg.cfg b/cfg/tarantool_silverbox_cfg.cfg
index 162e07540db3c17cfb22331f06a49e484d0008cb..de8d1edb3e11d37d50ef0f03a2871435303cb78c 100644
--- a/cfg/tarantool_silverbox_cfg.cfg
+++ b/cfg/tarantool_silverbox_cfg.cfg
@@ -96,6 +96,11 @@ local_hot_standby = 0
# delay in fractional seconds between successive re-readings of wal_dir
wal_dir_rescan_delay = 0.1
+# panic if where is error reading snap or wal
+# be default panic any snapshot reading error and ignore errors then reading wals
+panic_on_snap_error = 1
+panic_on_wal_error = 0
+
# Remote hot standby (if enabled server will run in hot standby mode
# continuously fetching WAL records from wal_feeder_ipaddr:wal_feeder_port
remote_hot_standby = 0
diff --git a/cfg/tarantool_silverbox_cfg.cfg_tmpl b/cfg/tarantool_silverbox_cfg.cfg_tmpl
index b59c1a6e558ba8c2c01707059ca7773ada53ec55..444d681863f65c09089cae2923c0d7424b930969 100644
--- a/cfg/tarantool_silverbox_cfg.cfg_tmpl
+++ b/cfg/tarantool_silverbox_cfg.cfg_tmpl
@@ -80,7 +80,6 @@ memcached_expire_full_sweep=3600
# do not write snapshot faster then snap_io_rate_limit MBytes/sec
snap_io_rate_limit=0.0
-
# Write no more rows in WAL
rows_per_wal=500000
@@ -97,6 +96,12 @@ local_hot_standby=0
# delay in fractional seconds between successive re-readings of wal_dir
wal_dir_rescan_delay=0.1
+
+# panic if where is error reading snap or wal
+# be default panic any snapshot reading error and ignore errors then reading wals
+panic_on_snap_error=1
+panic_on_wal_error=0
+
# Remote hot standby (if enabled server will run in hot standby mode
# continuously fetching WAL records from wal_feeder_ipaddr:wal_feeder_port
remote_hot_standby=0
diff --git a/cfg/tarantool_silverbox_cfg.h b/cfg/tarantool_silverbox_cfg.h
index 5178a74ce4badf33ffb9cc24d15ae31687ae017d..02d6b6559ffb0ad9b22791cd0c42ce369220a10e 100644
--- a/cfg/tarantool_silverbox_cfg.h
+++ b/cfg/tarantool_silverbox_cfg.h
@@ -135,6 +135,13 @@ typedef struct tarantool_cfg {
/* delay in fractional seconds between successive re-readings of wal_dir */
double wal_dir_rescan_delay;
+ /*
+ * panic if where is error reading snap or wal
+ * be default panic any snapshot reading error and ignore errors then reading wals
+ */
+ int32_t panic_on_snap_error;
+ int32_t panic_on_wal_error;
+
/*
* Remote hot standby (if enabled server will run in hot standby mode
* continuously fetching WAL records from wal_feeder_ipaddr:wal_feeder_port
diff --git a/core/log_io.c b/core/log_io.c
index 8061104cc30e3b52a9586fbb711086459e02a0b1..7dfd45ca2da3c9c43d51eb8d4582869b485e64fb 100644
--- a/core/log_io.c
+++ b/core/log_io.c
@@ -266,6 +266,8 @@ read_rows(struct log_io_iter *i)
goto eof;
if (row == NULL) {
+ if (l->class->panic_if_error)
+ panic("failed to read row");
say_warn("failed to read row");
goto restart;
}
@@ -1305,6 +1307,18 @@ recover_init(const char *snap_dirname, const char *wal_dirname,
return r;
}
+void
+recovery_setup_panic(struct recovery_state *r, bool on_snap_error, bool on_wal_error)
+{
+ struct log_io_class **class;
+
+ for (class = r->wal_class; *class; class++)
+ (*class)->panic_if_error = on_wal_error;
+
+ for (class = r->snap_class; *class; class++)
+ (*class)->panic_if_error = on_snap_error;
+}
+
static void
write_rows(struct log_io_iter *i)
{
diff --git a/include/log_io.h b/include/log_io.h
index 02f75308a6f8d4d341580a443083606d6ce0f4d8..52398070d16cdbc9662eab180cd54ba18de60143 100644
--- a/include/log_io.h
+++ b/include/log_io.h
@@ -55,6 +55,7 @@ struct log_io_class {
size_t marker_size, eof_marker_size;
size_t rows_per_file;
double fsync_delay;
+ bool panic_if_error;
const char *filetype;
const char *version;
@@ -124,6 +125,8 @@ void recover_follow(struct recovery_state *r, ev_tstamp wal_dir_rescan_delay);
void recover_finalize(struct recovery_state *r);
bool wal_write(struct recovery_state *r, i64 lsn, struct tbuf *data);
+void recovery_setup_panic(struct recovery_state *r, bool on_snap_error, bool on_wal_error);
+
/* recovery accessors */
struct palloc_pool *recovery_pool(struct recovery_state *r);
int confirm_lsn(struct recovery_state *r, i64 lsn);
diff --git a/mod/feeder/feeder.c b/mod/feeder/feeder.c
index 04ff9f265a87c5f78ba71d3727eba01ebac7dda9..05c2a5dd32def3ff7da4c2d1abfaaa2808e5b5ca 100644
--- a/mod/feeder/feeder.c
+++ b/mod/feeder/feeder.c
@@ -80,7 +80,7 @@ recover_feed_slave(int sock)
send_row(NULL, ver);
log_io = recover_init(NULL, cfg.wal_feeder_dir,
- NULL, NULL, send_row, 0, 0, 0, 64, RECOVER_READONLY, false);
+ NULL, NULL, send_row, 0, 0, 64, RECOVER_READONLY, false);
recover(log_io, lsn);
recover_follow(log_io, 0.1);
diff --git a/mod/silverbox/box.c b/mod/silverbox/box.c
index c3e881c0965d0718024328cb350b6a9bf0878059..802f4f5b46e5d93c985246bec042e0b5dc90b386 100644
--- a/mod/silverbox/box.c
+++ b/mod/silverbox/box.c
@@ -1775,6 +1775,7 @@ mod_init(void)
init_storage ? RECOVER_READONLY : 0, NULL);
recovery_state->snap_io_rate_limit = cfg.snap_io_rate_limit * 1024 * 1024;
+ recovery_setup_panic(recovery_state, cfg.panic_on_snap_error, cfg.panic_on_wal_error);
/* initialize hashes _after_ starting wal writer */
if (cfg.memcached != 0) {
diff --git a/mod/silverbox/box_cfg.cfg_tmpl b/mod/silverbox/box_cfg.cfg_tmpl
index c8d8d2544d0dfc60c51507e60fcfed8caa0b7ba5..1bc8a9bf9d4a9b1a7d8b035dd898384057fa5551 100644
--- a/mod/silverbox/box_cfg.cfg_tmpl
+++ b/mod/silverbox/box_cfg.cfg_tmpl
@@ -32,7 +32,6 @@ memcached_expire_full_sweep=3600
# do not write snapshot faster then snap_io_rate_limit MBytes/sec
snap_io_rate_limit=0.0
-
# Write no more rows in WAL
rows_per_wal=500000
@@ -49,6 +48,12 @@ local_hot_standby=0
# delay in fractional seconds between successive re-readings of wal_dir
wal_dir_rescan_delay=0.1
+
+# panic if where is error reading snap or wal
+# be default panic any snapshot reading error and ignore errors then reading wals
+panic_on_snap_error=1
+panic_on_wal_error=0
+
# Remote hot standby (if enabled server will run in hot standby mode
# continuously fetching WAL records from wal_feeder_ipaddr:wal_feeder_port
remote_hot_standby=0