Branch: refs/heads/master
Home: https://github.com/mailru/tarantool
Commit: 9120420807345fe235143e4130c1dac14bbcfe9c
https://github.com/mailru/tarantool/commit/9120420807345fe235143e4130c1dac14bbcfe9c
Author: Konstantin Osipov <[email protected]>
Date: 2012-05-03 (Thu, 03 May 2012)
Changed paths:
M include/log_io.h
M mod/box/box.m
M src/log_io.m
M src/replication.m
Log Message:
-----------
Add comments to log_io.m. Fix style. Renames.
diff --git a/include/log_io.h b/include/log_io.h
index e611b19..6a0f731 100644
--- a/include/log_io.h
+++ b/include/log_io.h
@@ -186,8 +186,8 @@ void recovery_init(const char *snap_dirname, const char
*xlog_dirname,
void recovery_update_io_rate_limit(double new_limit);
void recovery_free();
int recover(struct recovery_state *, i64 lsn);
-void recover_follow(struct recovery_state *r, ev_tstamp wal_dir_rescan_delay);
-void recover_finalize(struct recovery_state *r);
+void recovery_follow_local(struct recovery_state *r, ev_tstamp
wal_dir_rescan_delay);
+void recovery_finalize(struct recovery_state *r);
int wal_write(struct recovery_state *r, u16 tag, u16 op,
u64 cookie, i64 lsn, struct tbuf *data);
diff --git a/mod/box/box.m b/mod/box/box.m
index b61f531..d90f1de 100644
--- a/mod/box/box.m
+++ b/mod/box/box.m
@@ -1844,7 +1844,7 @@ void txn_assign_n(struct box_txn *txn, struct tbuf *data)
static void
box_leave_local_standby_mode(void *data __attribute__((unused)))
{
- recover_finalize(recovery_state);
+ recovery_finalize(recovery_state);
box_enter_master_or_replica_mode(&cfg);
}
@@ -2168,7 +2168,7 @@ void txn_assign_n(struct box_txn *txn, struct tbuf *data)
if (cfg.local_hot_standby) {
say_info("starting local hot standby");
- recover_follow(recovery_state, cfg.wal_dir_rescan_delay);
+ recovery_follow_local(recovery_state, cfg.wal_dir_rescan_delay);
snprintf(status, sizeof(status), "hot_standby");
title("hot_standby");
}
diff --git a/src/log_io.m b/src/log_io.m
index 31d2014..8d1809d 100644
--- a/src/log_io.m
+++ b/src/log_io.m
@@ -45,6 +45,67 @@
#include <crc32.h>
#include <tarantool_pthread.h>
#include "errinj.h"
+/*
+ * Recovery subsystem
+ * ------------------
+ *
+ * A facade of the recovery subsystem is struct recovery_state,
+ * which is a singleton.
+ *
+ * Depending on the configuration, start-up parameters, the
+ * actual task being performed, the recovery can be
+ * in a different state.
+ *
+ * The main factors influencing recovery state are:
+ * - temporal: whether or not the instance is just booting
+ * from a snapshot, is in 'local hot standby mode', or
+ * is already accepting requests
+ * - topological: whether or not it is a master instance
+ * or a replica
+ * - task based: whether it's a master process,
+ * snapshot saving process or a replication relay.
+ *
+ * Depending on the above factors, recovery can be in two main
+ * operation modes: "read mode", recovering in-memory state
+ * from existing data, and "write mode", i.e. recording on
+ * disk changes of the in-memory state.
+ *
+ * Let's enumerate all possible distinct states of recovery:
+ *
+ * Read mode
+ * ---------
+ * IR - initial recovery, initiated right after server start:
+ * reading data from the snapshot and existing WALs
+ * and restoring the in-memory state
+ * IRR - initial replication relay mode, reading data from
+ * existing WALs (xlogs) and sending it to the client.
+ *
+ * HS - standby mode, entered once all existing WALs are read:
+ * following the WAL directory for all changes done by the master
+ * and updating the in-memory state
+ * RR - replication relay, following the WAL directory for all
+ * changes done by the master and sending them to the
+ * replica
+ *
+ * Write mode
+ * ----------
+ * M - master mode, recording in-memory state changes in the WAL
+ * R - replica mode, receiving changes from the master and
+ * recording them in the WAL
+ * S - snapshot mode, writing entire in-memory state to a compact
+ * snapshot file.
+ *
+ * The following state transitions are possible/supported:
+ *
+ * recovery_init() -> IR | IRR # recover()
+ * IR -> HS # recovery_follow_local()
+ * IRR -> RR # recovery_follow_local()
+ * HS -> M # recovery_finalize()
+ * M -> R # recovery_follow_remote()
+ * R -> M # recovery_stop_remote()
+ * M -> S # snapshot()
+ * R -> S # snapshot()
+ */
const u16 snap_tag = -1;
const u16 wal_tag = -2;
@@ -698,7 +759,9 @@
static struct log_io *
-log_io_open_for_read(struct recovery_state *recover, struct log_io_class
*class, i64 lsn, int suffix,
+log_io_open_for_read(struct recovery_state *recover,
+ struct log_io_class *class,
+ i64 lsn, int suffix,
const char *filename)
{
const char *errmsg;
@@ -721,7 +784,7 @@
strncpy(l->filename, filename, PATH_MAX);
}
- say_debug("%s: opening %s'", __func__, l->filename);
+ say_debug("%s: opening %s", __func__, l->filename);
l->f = fopen(l->filename, "r");
if (l->f == NULL) {
@@ -998,7 +1061,10 @@
goto recover_current_wal;
while (r->confirmed_lsn < wal_greatest_lsn) {
- /* if newer WAL appeared in directory before current_wal was
fully read try reread last */
+ /*
+ * If a newer WAL appeared in the directory before
+ * current_wal was fully read, try re-reading
+ * one last time. */
if (r->current_wal != NULL) {
if (r->current_wal->retry++ < 3) {
say_warn("try reread `%s' despite newer WAL
exists",
@@ -1011,9 +1077,10 @@
}
}
- current_lsn = r->confirmed_lsn + 1; /* TODO: find better
way looking for next xlog */
- next_wal = log_io_open_for_read(r, r->wal_class, current_lsn,
0, NULL);
-
+ /* TODO: find a better way of finding the next xlog */
+ current_lsn = r->confirmed_lsn + 1;
+ next_wal = log_io_open_for_read(r, r->wal_class, current_lsn,
+ 0, NULL);
/*
* When doing final recovery, and dealing with the
* last file, try opening .<suffix>.inprogress.
@@ -1035,30 +1102,33 @@
break;
}
-
assert(r->current_wal == NULL);
r->current_wal = next_wal;
say_info("recover from `%s'", r->current_wal->filename);
- recover_current_wal:
+recover_current_wal:
rows_before = r->current_wal->rows;
result = recover_wal(r, r->current_wal);
if (result < 0) {
- say_error("failure reading from %s",
r->current_wal->filename);
+ say_error("failure reading from %s",
+ r->current_wal->filename);
break;
}
- if (r->current_wal->rows > 0 && r->current_wal->rows !=
rows_before)
+ if (r->current_wal->rows > 0 &&
+ r->current_wal->rows != rows_before) {
r->current_wal->retry = 0;
-
- /* rows == 0 could possible indicate to an empty WAL */
+ }
+ /* rows == 0 could indicate an empty WAL */
if (r->current_wal->rows == 0) {
- say_error("read zero records from %s",
r->current_wal->filename);
+ say_error("read zero records from %s",
+ r->current_wal->filename);
break;
}
if (result == LOG_EOF) {
- say_info("done `%s' confirmed_lsn:%" PRIi64,
r->current_wal->filename,
+ say_info("done `%s' confirmed_lsn:%" PRIi64,
+ r->current_wal->filename,
r->confirmed_lsn);
log_io_close(&r->current_wal);
}
@@ -1133,10 +1203,10 @@
return result;
}
-static void recover_follow_file(ev_stat *w, int revents
__attribute__((unused)));
+static void recovery_follow_file(ev_stat *w, int revents
__attribute__((unused)));
static void
-recover_follow_dir(ev_timer *w, int revents __attribute__((unused)))
+recovery_follow_dir(ev_timer *w, int revents __attribute__((unused)))
{
struct recovery_state *r = w->data;
struct log_io *wal = r->current_wal;
@@ -1147,13 +1217,13 @@
/* recover_remaining_wals found new wal */
if (r->current_wal != NULL && wal != r->current_wal) {
ev_stat *stat = &r->current_wal->stat;
- ev_stat_init(stat, recover_follow_file,
r->current_wal->filename, 0.);
+ ev_stat_init(stat, recovery_follow_file,
r->current_wal->filename, 0.);
ev_stat_start(stat);
}
}
static void
-recover_follow_file(ev_stat *w, int revents __attribute__((unused)))
+recovery_follow_file(ev_stat *w, int revents __attribute__((unused)))
{
struct recovery_state *r = w->data;
int result;
@@ -1161,28 +1231,29 @@
if (result < 0)
panic("recover failed");
if (result == LOG_EOF) {
- say_info("done `%s' confirmed_lsn:%" PRIi64,
r->current_wal->filename,
+ say_info("done `%s' confirmed_lsn:%" PRIi64,
+ r->current_wal->filename,
r->confirmed_lsn);
log_io_close(&r->current_wal);
- recover_follow_dir((ev_timer *)w, 0);
+ recovery_follow_dir((ev_timer *)w, 0);
}
}
void
-recover_follow(struct recovery_state *r, ev_tstamp wal_dir_rescan_delay)
+recovery_follow_local(struct recovery_state *r, ev_tstamp wal_dir_rescan_delay)
{
- ev_timer_init(&r->wal_timer, recover_follow_dir,
+ ev_timer_init(&r->wal_timer, recovery_follow_dir,
wal_dir_rescan_delay, wal_dir_rescan_delay);
ev_timer_start(&r->wal_timer);
if (r->current_wal != NULL) {
ev_stat *stat = &r->current_wal->stat;
- ev_stat_init(stat, recover_follow_file,
r->current_wal->filename, 0.);
+ ev_stat_init(stat, recovery_follow_file,
r->current_wal->filename, 0.);
ev_stat_start(stat);
}
}
void
-recover_finalize(struct recovery_state *r)
+recovery_finalize(struct recovery_state *r)
{
int result;
diff --git a/src/replication.m b/src/replication.m
index efc782d..c58c368 100644
--- a/src/replication.m
+++ b/src/replication.m
@@ -560,7 +560,6 @@ static void spawner_signal_handler(int signal)
{
char name[FIBER_NAME_MAXLEN];
struct sigaction sa;
- struct recovery_state *log_io;
struct tbuf *ver;
i64 lsn;
ssize_t r;
@@ -619,10 +618,8 @@ static void spawner_signal_handler(int signal)
INT32_MAX, "fsync_delay", 0,
RECOVER_READONLY, false);
- log_io = recovery_state;
-
- recover(log_io, lsn);
- recover_follow(log_io, 0.1);
+ recover(recovery_state, lsn);
+ recovery_follow_local(recovery_state, 0.1);
ev_loop(0);
================================================================
_______________________________________________
Mailing list: https://launchpad.net/~tarantool-developers
Post to : [email protected]
Unsubscribe : https://launchpad.net/~tarantool-developers
More help : https://help.launchpad.net/ListHelp
