hcq/oceanbase
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

add sys_recovery_scn to __all_tenant_info && primary tenant iter log

Browse Source
This commit is contained in:
maosy
2023-02-27 10:41:19 +00:00
committed by ob-robot
parent 41c067141f
commit 09f099c4e2
55 changed files with 4347 additions and 3733 deletions
Download Patch File Download Diff File Expand all files Collapse all files

657
src/rootserver/ob_recovery_ls_service.cpp
View File

@ -20,28 +20,16 @@
#include "logservice/ob_log_handler.h" //ObLogHandler
#include "logservice/palf/log_entry.h" //LogEntry
#include "logservice/palf/log_define.h"
#include "logservice/ob_log_service.h"
#include "share/scn.h"//SCN
#include "logservice/ob_garbage_collector.h"//ObGCLSLog
#include "logservice/restoreservice/ob_log_restore_handler.h"//ObLogRestoreHandler
#include "observer/ob_server_struct.h" //GCTX
#include "rootserver/ob_primary_ls_service.h" //ObTenantLSInfo
#include "rootserver/ob_tenant_recovery_reportor.h" //ObTenantRecoveryReportor
#include "rootserver/ob_tenant_info_loader.h" // ObTenantInfoLoader
#include "share/ls/ob_ls_life_manager.h" //ObLSLifeManger
#include "share/ls/ob_ls_operator.h" //ObLSAttr
#include "share/ls/ob_ls_recovery_stat_operator.h" //ObLSRecoveryLSStatOperator
#include "share/ob_errno.h"
#include "share/ob_share_util.h" //ObShareUtil
#include "share/schema/ob_multi_version_schema_service.h" //ObMultiSchemaService
#include "share/restore/ob_physical_restore_info.h" //restore_status
#include "share/restore/ob_physical_restore_table_operator.h"//ObPhysicalRestoreTableOperator
#include "share/ob_primary_standby_service.h" // ObPrimaryStandbyService
#include "share/ob_standby_upgrade.h" // ObStandbyUpgrade
#include "share/ob_upgrade_utils.h" // ObUpgradeChecker
#include "share/ob_global_stat_proxy.h" // ObGlobalStatProxy
#include "storage/tx/ob_tx_log.h" //ObTxLogHeader
#include "storage/tx_storage/ob_ls_service.h" //ObLSService
#include "storage/tx_storage/ob_ls_handle.h" //ObLSHandle
#include "share/ob_cluster_version.h"//GET_TENANT_DATA_VERSION
#include "rootserver/ob_ls_service_helper.h"//ObRecoveryLSHelper
namespace oceanbase
{
@ -64,7 +52,7 @@ int ObRecoveryLSService::init()
ret = OB_INIT_TWICE;
LOG_WARN("has inited", KR(ret));
} else if (OB_FAIL(ObTenantThreadHelper::create("RecLSSer",
lib::TGDefIDs::LSService, *this))) {
lib::TGDefIDs::SimpleLSService, *this))) {
LOG_WARN("failed to create thread", KR(ret));
} else if (OB_FAIL(ObTenantThreadHelper::start())) {
LOG_WARN("failed to start thread", KR(ret));
@ -92,318 +80,40 @@ void ObRecoveryLSService::do_work()
if (OB_UNLIKELY(!inited_) || OB_ISNULL(proxy_)) {
ret = OB_NOT_INIT;
LOG_WARN("not init", K(ret), K(inited_), KP(proxy_));
} else if (OB_FAIL(wait_tenant_data_version_ready_(tenant_id_, DATA_VERSION_4_1_0_0))) {
LOG_WARN("failed to wait tenant and version ready", KR(ret), K(tenant_id_));
} else {
ObLSRecoveryStatOperator ls_recovery;
palf::PalfBufferIterator iterator;
int tmp_ret = OB_SUCCESS;
const int64_t idle_time_us = 100 * 1000L;
int64_t idle_time_us = 100 * 1000L;
SCN start_scn;
while (!has_set_stop()) {
ObCurTraceId::init(GCONF.self_addr_);
uint64_t thread_idx = get_thread_idx();
ObTenantInfoLoader *tenant_info_loader = MTL(ObTenantInfoLoader*);
ObAllTenantInfo tenant_info;
//two thread for seed log and recovery_ls_manager
// two thread for seed log and recovery_ls_manager
if (!is_user_tenant(tenant_id_)) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("ls recovery thread must run on user tenant", KR(ret), K(tenant_id_));
} else if (OB_ISNULL(tenant_info_loader)) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("tenant report is null", KR(ret), K(tenant_id_));
} else if (OB_FAIL(tenant_info_loader->get_tenant_info(tenant_info))) {
LOG_WARN("failed to get tenant info", KR(ret));
} else if (OB_FAIL(check_can_do_recovery_(tenant_info))) {
LOG_WARN("failed to check do recovery", KR(ret), K(tenant_info));
} else if (0 == thread_idx) {
if (OB_SUCCESS != (tmp_ret = process_recovery_ls_manager())) {
ret = OB_SUCC(ret) ? tmp_ret : ret;
LOG_WARN("failed to process recovery ls manager", KR(ret), KR(tmp_ret));
}
} else {
if (!start_scn.is_valid()) {
ObLSRecoveryStat ls_recovery_stat;
if (OB_FAIL(ls_recovery.get_ls_recovery_stat(tenant_id_,
SYS_LS, false, ls_recovery_stat, *proxy_))) {
LOG_WARN("failed to load sys recovery stat", KR(ret), K(tenant_id_));
} else if (SCN::base_scn() == ls_recovery_stat.get_sync_scn()) {
//等待物理恢复设置完系统日志流初始同步点之后开始迭代
ret = OB_EAGAIN;
LOG_WARN("restore init sync scn has not setted, need wait", KR(ret), K(ls_recovery_stat));
} else if (OB_FAIL(report_sys_ls_recovery_stat_(ls_recovery_stat.get_sync_scn()))) {
//may recovery end, but readable scn need report
LOG_WARN("failed to report ls recovery stat", KR(ret), K(ls_recovery_stat));
} else if (tenant_info.get_recovery_until_scn() == ls_recovery_stat.get_sync_scn()) {
ret = OB_EAGAIN;
if (REACH_TIME_INTERVAL(60 * 1000 * 1000)) { // every minute
LOG_INFO("has recovered to recovery_until_scn", KR(ret), K(ls_recovery_stat), K(tenant_info));
}
} else if (OB_FAIL(seek_log_iterator_(ls_recovery_stat.get_sync_scn(), iterator))) {
LOG_WARN("failed to seek log iterator", KR(ret), K(ls_recovery_stat));
} else {
start_scn = ls_recovery_stat.get_sync_scn();
LOG_INFO("start to seek at", K(start_scn));
}
ObRecoveryLSHelper recovery_ls(tenant_id_, proxy_);
if (OB_FAIL(recovery_ls.do_work(iterator, start_scn))) {
LOG_WARN("failed to recovery ls", KR(ret), K(start_scn));
}
if (OB_FAIL(ret)) {
} else if (start_scn.is_valid() && OB_FAIL(process_ls_log_(tenant_info, start_scn, iterator))) {
if (OB_ITER_STOP != ret) {
LOG_WARN("failed to process ls log", KR(ret), K(start_scn), K(tenant_info));
}
}
if (OB_FAIL(ret)) {
start_scn.reset();
}
if (REACH_TIME_INTERVAL(10 * 1000 * 1000)) { // every 10 second
(void)try_tenant_upgrade_end_();
if (OB_ITER_STOP == ret) {
//standby or recovery may restore finish,
//so when the error code is OB_ITER_STOP, adjust idle_time to 10s
idle_time_us = 10 * 1000 * 1000;//10s
}
}
LOG_INFO("[LS_RECOVERY] finish one round", KR(ret), KR(tmp_ret), K(start_scn), K(thread_idx), K(tenant_info));
if (REACH_TIME_INTERVAL(10 * 1000 * 1000)) { // every 10 second
(void)try_tenant_upgrade_end_();
}
LOG_INFO("[RECOVERY_LS_SERVICE] finish one round", KR(ret),
K(start_scn));
idle(idle_time_us);
}
}
}
int ObRecoveryLSService::seek_log_iterator_(const SCN &sync_scn, PalfBufferIterator &iterator)
{
int ret = OB_SUCCESS;
if (OB_UNLIKELY(!inited_)) {
ret = OB_NOT_INIT;
LOG_WARN("not init", KR(ret), K(inited_));
} else if (OB_UNLIKELY(!sync_scn.is_valid())) {
ret = OB_INVALID_ARGUMENT;
LOG_WARN("sync scn is invalid", KR(ret), K(sync_scn));
} else {
ObLSService *ls_svr = MTL(ObLSService *);
ObLSHandle ls_handle;
if (OB_FAIL(ls_svr->get_ls(SYS_LS, ls_handle, storage::ObLSGetMod::RS_MOD))) {
LOG_WARN("failed to get ls", KR(ret));
} else {
ObLogHandler *log_handler = NULL;
ObLS *ls = NULL;
palf::LSN start_lsn;
if (OB_ISNULL(ls = ls_handle.get_ls())
|| OB_ISNULL(log_handler = ls->get_log_handler())) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("ls or log handle is null", KR(ret), KP(ls),
KP(log_handler));
} else if (OB_FAIL(log_handler->locate_by_scn_coarsely(sync_scn, start_lsn))) {
LOG_WARN("failed to locate lsn", KR(ret), K(sync_scn));
} else if (OB_FAIL(log_handler->seek(start_lsn, iterator))) {
LOG_WARN("failed to seek iterator", KR(ret), K(sync_scn));
}
}
}
return ret;
}
int ObRecoveryLSService::process_ls_log_(
const ObAllTenantInfo &tenant_info,
SCN &start_scn,
PalfBufferIterator &iterator)
{
int ret = OB_SUCCESS;
palf::LogEntry log_entry;
palf::LSN target_lsn;
SCN sync_scn;
SCN last_sync_scn;
if (OB_UNLIKELY(!inited_)) {
ret = OB_NOT_INIT;
LOG_WARN("not init", KR(ret), K(inited_));
} else if (OB_UNLIKELY(!tenant_info.is_valid())) {
ret = OB_INVALID_ARGUMENT;
LOG_WARN("tenant_info is invalid", KR(ret), K(tenant_info));
}
while (OB_SUCC(ret) && OB_SUCC(iterator.next())) {
if (OB_FAIL(iterator.get_entry(log_entry, target_lsn))) {
LOG_WARN("failed to get log", KR(ret), K(log_entry));
} else if (OB_ISNULL(log_entry.get_data_buf())) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("log entry is null", KR(ret));
} else {
LOG_DEBUG("get log", K(log_entry), K(target_lsn), K(start_scn));
sync_scn = log_entry.get_scn();
const char *log_buf = log_entry.get_data_buf();
const int64_t log_length = log_entry.get_data_len();
logservice::ObLogBaseHeader header;
const int64_t HEADER_SIZE = header.get_serialize_size();
int64_t log_pos = 0;
if (OB_UNLIKELY(sync_scn <= start_scn)) {
//通过scn定位的LSN是不准确的，可能会获取多余的数据，所以需要把小于等于sync_scn的过滤掉
continue;
} else if (tenant_info.get_recovery_until_scn() < sync_scn) {
if (OB_FAIL(report_sys_ls_recovery_stat_(tenant_info.get_recovery_until_scn()))) {
LOG_WARN("failed to report_sys_ls_recovery_stat_", KR(ret), K(sync_scn), K(tenant_info),
K(log_entry), K(target_lsn), K(start_scn));
// SYS LS has recovered to the recovery_until_scn, need stop iterate SYS LS log and reset start_scn
} else if (OB_FAIL(seek_log_iterator_(tenant_info.get_recovery_until_scn(), iterator))) {
LOG_WARN("failed to seek log iterator", KR(ret), K(sync_scn), K(tenant_info),
K(log_entry), K(target_lsn), K(start_scn));
} else {
ret = OB_ITER_STOP;
LOG_WARN("SYS LS has recovered to the recovery_until_scn, need stop iterate SYS LS log",
KR(ret), K(sync_scn), K(tenant_info), K(log_entry), K(target_lsn), K(start_scn));
start_scn.reset();
}
} else if (OB_FAIL(header.deserialize(log_buf, HEADER_SIZE, log_pos))) {
LOG_WARN("failed to deserialize", KR(ret), K(HEADER_SIZE));
} else if (OB_UNLIKELY(log_pos >= log_length)) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("log pos is not expected", KR(ret), K(log_pos), K(log_length));
} else if (logservice::GC_LS_LOG_BASE_TYPE == header.get_log_type()) {
ObGCLSLog gc_log;
if (OB_FAIL(gc_log.deserialize(log_buf, log_length, log_pos))) {
LOG_WARN("failed to deserialize gc log", KR(ret), K(log_length));
} else if (ObGCLSLOGType::OFFLINE_LS == gc_log.get_log_type()) {
//set sys ls offline
if (OB_FAIL(process_gc_log_(gc_log, sync_scn))) {
LOG_WARN("failed to process gc log", KR(ret), K(sync_scn));
}
}
// nothing
} else if (logservice::TRANS_SERVICE_LOG_BASE_TYPE == header.get_log_type()) {
ObTxLogBlock tx_log_block;
ObTxLogBlockHeader tx_block_header;
if (OB_FAIL(tx_log_block.init(log_buf, log_length, log_pos, tx_block_header))) {
LOG_WARN("failed to init tx log block", KR(ret), K(log_length));
} else if (OB_FAIL(process_ls_tx_log_(tx_log_block, sync_scn))) {
LOG_WARN("failed to process ls tx log", KR(ret), K(tx_log_block), K(sync_scn));
}
} else {}
if (OB_SUCC(ret)) {
last_sync_scn = sync_scn;
} else if (last_sync_scn.is_valid()) {
//if ls_operator can not process, need to report last sync scn
int tmp_ret = OB_SUCCESS;
if (OB_TMP_FAIL(report_sys_ls_recovery_stat_(last_sync_scn))) {
LOG_WARN("failed to report ls recovery stat", KR(ret), KR(tmp_ret), K(last_sync_scn));
}
}
}
}//end for each log
if (OB_ITER_END == ret) {
ret = OB_SUCCESS;
if (OB_FAIL(report_sys_ls_recovery_stat_(sync_scn))) {
LOG_WARN("failed to report ls recovery stat", KR(ret), K(sync_scn));
}
} else if (OB_SUCC(ret)) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("iterator must be end", KR(ret));
} else {
LOG_WARN("failed to get next log", KR(ret));
}
return ret;
}
int ObRecoveryLSService::process_ls_tx_log_(ObTxLogBlock &tx_log_block, const SCN &sync_scn)
{
int ret = OB_SUCCESS;
bool has_operation = false;
if (OB_UNLIKELY(!inited_)) {
ret = OB_NOT_INIT;
LOG_WARN("not init", KR(ret), K(inited_));
}
while (OB_SUCC(ret)) {
transaction::ObTxLogHeader tx_header;
if (OB_FAIL(tx_log_block.get_next_log(tx_header))) {
if (OB_ITER_END == ret) {
ret = OB_SUCCESS;
break;
} else {
LOG_WARN("failed to get next log", KR(ret));
}
} else if (transaction::ObTxLogType::TX_COMMIT_LOG !=
tx_header.get_tx_log_type()) {
// nothing
} else {
ObTxCommitLogTempRef temp_ref;
ObTxCommitLog commit_log(temp_ref);
const int64_t COMMIT_SIZE = commit_log.get_serialize_size();
//TODO commit log may too large
if (OB_FAIL(tx_log_block.deserialize_log_body(commit_log))) {
LOG_WARN("failed to deserialize", KR(ret));
} else {
const ObTxBufferNodeArray &source_data =
commit_log.get_multi_source_data();
for (int64_t i = 0; OB_SUCC(ret) && i < source_data.count(); ++i) {
const ObTxBufferNode &node = source_data.at(i);
if (ObTxDataSourceType::STANDBY_UPGRADE == node.get_data_source_type()) {
if (OB_FAIL(process_upgrade_log_(node))) {
LOG_WARN("failed to process_upgrade_log_", KR(ret), K(node));
}
} else if (ObTxDataSourceType::LS_TABLE != node.get_data_source_type()) {
// nothing
} else if (has_operation) {
ret = OB_ERR_UNEXPECTED;
LOG_ERROR("one clog has more than one operation", KR(ret), K(commit_log));
} else {
has_operation = true;
ObLSAttr ls_attr;
const int64_t LS_SIZE = ls_attr.get_serialize_size();
int64_t pos = 0;
if (OB_FAIL(ls_attr.deserialize(node.get_data_buf().ptr(), LS_SIZE,
pos))) {
LOG_WARN("failed to deserialize", KR(ret), K(node), K(LS_SIZE));
} else if (OB_UNLIKELY(pos > LS_SIZE)) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("failed to get ls attr", KR(ret), K(pos), K(LS_SIZE));
} else {
LOG_INFO("get ls operation", K(ls_attr), K(sync_scn));
//TODO ls recovery is too fast for ls manager, so it maybe failed, while change ls status
//consider how to retry
if (OB_FAIL(process_ls_operator_(ls_attr,
sync_scn))) {
LOG_WARN("failed to process ls operator", KR(ret), K(ls_attr),
K(sync_scn));
}
}
}
}// end for
}
}
} // end while for each tx_log
return ret;
}
int ObRecoveryLSService::process_upgrade_log_(const ObTxBufferNode &node)
{
int ret = OB_SUCCESS;
uint64_t standby_data_version = 0;
if (!node.is_valid()) {
ret = OB_INVALID_ARGUMENT;
LOG_WARN("node is invalid", KR(ret), K(node));
} else {
ObStandbyUpgrade primary_data_version;
int64_t pos = 0;
if (OB_FAIL(primary_data_version.deserialize(node.get_data_buf().ptr(), node.get_data_buf().length(), pos))) {
LOG_WARN("failed to deserialize", KR(ret), K(node), KPHEX(node.get_data_buf().ptr(), node.get_data_buf().length()));
} else if (OB_UNLIKELY(pos > node.get_data_buf().length())) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("failed to get primary_data_version", KR(ret), K(pos), K(node.get_data_buf().length()));
} else {
LOG_INFO("get primary_data_version", K(primary_data_version));
if (!primary_data_version.is_valid()) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("primary_data_version not valid", KR(ret), K(primary_data_version));
} else if (OB_FAIL(ObUpgradeChecker::get_data_version_by_cluster_version(
GET_MIN_CLUSTER_VERSION(),
standby_data_version))) {
LOG_WARN("failed to get_data_version_by_cluster_version", KR(ret), K(GET_MIN_CLUSTER_VERSION()));
} else if (primary_data_version.get_data_version() > standby_data_version) {
ret = OB_EAGAIN;
if (REACH_TIME_INTERVAL(30 * 60 * 1000 * 1000)) { // 30min
LOG_ERROR("standby version is not new enough to recover primary clog", KR(ret),
K(primary_data_version), K(standby_data_version));
}
}
}
}
return ret;
}
int ObRecoveryLSService::get_min_data_version_(uint64_t &compatible)
{
int ret = OB_SUCCESS;
@ -506,331 +216,6 @@ void ObRecoveryLSService::try_tenant_upgrade_end_()
}
}
int ObRecoveryLSService::process_gc_log_(logservice::ObGCLSLog &gc_log, const SCN &sync_scn)
{
int ret = OB_SUCCESS;
common::ObMySQLTransaction trans;
const uint64_t meta_tenant_id = gen_meta_tenant_id(tenant_id_);
ObLSLifeAgentManager ls_life_agent(*proxy_);
if (OB_ISNULL(proxy_)) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("proxy is null", KR(ret));
} else if (OB_FAIL(trans.start(proxy_, meta_tenant_id))) {
LOG_WARN("failed to start trans", KR(ret), K(meta_tenant_id));
} else if (OB_FAIL(ls_life_agent.set_ls_offline_in_trans(
tenant_id_, SYS_LS, share::OB_LS_TENANT_DROPPING, sync_scn, share::NORMAL_SWITCHOVER_STATUS,
trans))) {
LOG_WARN("failed to set offline", KR(ret), K(tenant_id_), K(sync_scn));
} else {
ObLSRecoveryStatOperator ls_recovery;
ObLSRecoveryStat ls_recovery_stat;
if (OB_FAIL(construct_ls_recovery_stat(sync_scn, ls_recovery_stat))) {
LOG_WARN("failed to construct ls recovery stat", KR(ret), K(sync_scn));
} else if (OB_FAIL(ls_recovery.update_ls_recovery_stat_in_trans(
ls_recovery_stat, trans))) {
LOG_WARN("failed to update ls recovery stat", KR(ret), K(ls_recovery_stat));
}
}
if (trans.is_started()) {
int tmp_ret = OB_SUCCESS;
if (OB_SUCCESS != (tmp_ret = trans.end(OB_SUCC(ret)))) {
ret = OB_SUCC(ret) ? tmp_ret : ret;
LOG_WARN("failed to end trans", KR(ret), K(tmp_ret));
}
}
return ret;
}
int ObRecoveryLSService::construct_ls_recovery_stat(const SCN &sync_scn,
ObLSRecoveryStat &ls_stat)
{
int ret = OB_SUCCESS;
ls_stat.reset();
SCN readable_scn;
if (OB_UNLIKELY(!inited_)) {
ret = OB_NOT_INIT;
LOG_WARN("not init", KR(ret), K(inited_));
} else if (OB_FAIL(ObTenantRecoveryReportor::get_readable_scn(SYS_LS, readable_scn))) {
LOG_WARN("failed to get readable scn", KR(ret));
} else if (OB_FAIL(ls_stat.init_only_recovery_stat(
tenant_id_, SYS_LS, sync_scn, readable_scn))) {
LOG_WARN("failed to init ls recovery stat", KR(ret), K(tenant_id_),
K(sync_scn), K(readable_scn));
}
return ret;
}
int ObRecoveryLSService::process_ls_operator_(const share::ObLSAttr &ls_attr, const SCN &sync_scn)
{
int ret = OB_SUCCESS;
common::ObMySQLTransaction trans;
const uint64_t meta_tenant_id = gen_meta_tenant_id(tenant_id_);
if (OB_UNLIKELY(!ls_attr.is_valid())) {
ret = OB_INVALID_ARGUMENT;
LOG_WARN("ls attr is invalid", KR(ret), K(ls_attr));
} else if (OB_ISNULL(proxy_)) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("proxy is null", KR(ret));
} else if (OB_FAIL(check_valid_to_operator_ls_(ls_attr, sync_scn))) {
LOG_WARN("failed to check valid to operator ls", KR(ret), K(sync_scn), K(ls_attr));
} else if (OB_FAIL(trans.start(proxy_, meta_tenant_id))) {
LOG_WARN("failed to start trans", KR(ret), K(meta_tenant_id));
} else if (OB_FAIL(process_ls_operator_in_trans_(ls_attr, sync_scn, trans))) {
LOG_WARN("failed to process ls operator in trans", KR(ret), K(ls_attr), K(sync_scn));
}
if (OB_SUCC(ret)) {
ObLSRecoveryStat ls_recovery_stat;
ObLSRecoveryStatOperator ls_recovery;
if (OB_FAIL(construct_ls_recovery_stat(sync_scn, ls_recovery_stat))) {
LOG_WARN("failed to construct ls recovery stat", KR(ret), K(sync_scn));
} else if (OB_FAIL(ls_recovery.update_ls_recovery_stat_in_trans(
ls_recovery_stat, trans))) {
LOG_WARN("failed to update ls recovery stat", KR(ret), K(ls_recovery_stat));
}
}
if (trans.is_started()) {
int tmp_ret = OB_SUCCESS;
if (OB_SUCCESS != (tmp_ret = trans.end(OB_SUCC(ret)))) {
ret = OB_SUCC(ret) ? tmp_ret : ret;
LOG_WARN("failed to end trans", KR(ret), K(tmp_ret));
}
}
return ret;
}
int ObRecoveryLSService::check_valid_to_operator_ls_(
const share::ObLSAttr &ls_attr, const SCN &sync_scn)
{
int ret = OB_SUCCESS;
if (OB_UNLIKELY(!inited_)) {
ret = OB_NOT_INIT;
LOG_WARN("not init", KR(ret), K(inited_));
} else if (OB_ISNULL(proxy_)) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("proxy is null", KR(ret));
} else if (OB_UNLIKELY(!sync_scn.is_valid() || !ls_attr.is_valid())) {
ret = OB_INVALID_ARGUMENT;
LOG_WARN("syns scn is invalid", KR(ret), K(sync_scn), K(ls_attr));
} else if (share::is_ls_tenant_drop_pre_op(ls_attr.get_ls_operatin_type())) {
ret = OB_ITER_STOP;
LOG_WARN("can not process ls operator after tenant dropping", K(ls_attr));
} else if (share::is_ls_tenant_drop_op(ls_attr.get_ls_operatin_type())) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("ls recovery must stop while pre tenant dropping", KR(ret), K(ls_attr));
} else {
SCN user_scn;
ObLSRecoveryStatOperator ls_recovery;
if (OB_FAIL(ls_recovery.get_user_ls_sync_scn(tenant_id_, *proxy_, user_scn))) {
LOG_WARN("failed to get user scn", KR(ret), K(tenant_id_));
} else if (user_scn >= sync_scn) {
//other ls has larger sync scn, ls can operator
} else {
ret = OB_NEED_WAIT;
LOG_WARN("can not process ls operator, need wait other ls sync", KR(ret),
K(user_scn), K(sync_scn));
}
}
return ret;
}
int ObRecoveryLSService::process_ls_operator_in_trans_(
const share::ObLSAttr &ls_attr, const SCN &sync_scn, ObMySQLTransaction &trans)
{
int ret = OB_SUCCESS;
const uint64_t meta_tenant_id = gen_meta_tenant_id(tenant_id_);
if (OB_UNLIKELY(!ls_attr.is_valid())) {
ret = OB_INVALID_ARGUMENT;
LOG_WARN("ls attr is invalid", KR(ret), K(ls_attr));
} else if (OB_ISNULL(proxy_)) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("proxy is null", KR(ret));
} else {
ObLSStatusOperator ls_operator;
share::ObLSStatusInfo ls_status;
ObLSLifeAgentManager ls_life_agent(*proxy_);
if (OB_UNLIKELY(!ls_attr.is_valid())) {
ret = OB_INVALID_ARGUMENT;
LOG_WARN("ls attr is invalid", KR(ret), K(ls_attr));
} else if (share::is_ls_create_pre_op(ls_attr.get_ls_operatin_type())) {
//create new ls;
if (OB_FAIL(create_new_ls_(ls_attr, sync_scn, trans))) {
LOG_WARN("failed to create new ls", KR(ret), K(sync_scn), K(ls_attr));
}
} else if (share::is_ls_create_abort_op(ls_attr.get_ls_operatin_type())) {
if (OB_FAIL(ls_life_agent.drop_ls_in_trans(tenant_id_, ls_attr.get_ls_id(), share::NORMAL_SWITCHOVER_STATUS, trans))) {
LOG_WARN("failed to drop ls", KR(ret), K(tenant_id_), K(ls_attr));
}
} else if (share::is_ls_drop_end_op(ls_attr.get_ls_operatin_type())) {
if (OB_FAIL(ls_life_agent.set_ls_offline_in_trans(tenant_id_, ls_attr.get_ls_id(),
ls_attr.get_ls_status(), sync_scn, share::NORMAL_SWITCHOVER_STATUS, trans))) {
LOG_WARN("failed to set offline", KR(ret), K(tenant_id_), K(ls_attr), K(sync_scn));
}
} else {
ObLSStatus target_status = share::OB_LS_EMPTY;
if (OB_FAIL(ls_operator.get_ls_status_info(tenant_id_, ls_attr.get_ls_id(),
ls_status, trans))) {
LOG_WARN("failed to get ls status", KR(ret), K(tenant_id_), K(ls_attr));
} else if (ls_status.ls_is_creating()) {
ret = OB_EAGAIN;
LOG_WARN("ls not created, need wait", KR(ret), K(ls_status));
} else if (share::is_ls_create_end_op(ls_attr.get_ls_operatin_type())) {
// set ls to normal
target_status = share::OB_LS_NORMAL;
} else if (share::is_ls_tenant_drop_op(ls_attr.get_ls_operatin_type())) {
target_status = share::OB_LS_TENANT_DROPPING;
} else if (share::is_ls_drop_pre_op(ls_attr.get_ls_operatin_type())) {
target_status = share::OB_LS_DROPPING;
} else if (share::is_ls_tenant_drop_pre_op(ls_attr.get_ls_operatin_type())) {
target_status = share::OB_LS_PRE_TENANT_DROPPING;
} else {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("unexpected operation type", KR(ret), K(ls_attr));
}
if (FAILEDx(ls_operator.update_ls_status(tenant_id_, ls_attr.get_ls_id(),
ls_status.status_, target_status,
share::NORMAL_SWITCHOVER_STATUS, trans))) {
LOG_WARN("failed to update ls status", KR(ret), K(tenant_id_), K(ls_attr),
K(ls_status), K(target_status));
}
LOG_INFO("[LS_RECOVERY] update ls status", KR(ret), K(ls_attr), K(target_status));
}
}
return ret;
}
int ObRecoveryLSService::create_new_ls_(const share::ObLSAttr &ls_attr,
const SCN &sync_scn,
common::ObMySQLTransaction &trans)
{
int ret = OB_SUCCESS;
if (!share::is_ls_create_pre_op(ls_attr.get_ls_operatin_type())) {
ret = OB_INVALID_ARGUMENT;
LOG_WARN("ls not create pre operation", KR(ret), K(ls_attr));
} else {
//create new ls;
share::schema::ObSchemaGetterGuard schema_guard;
const share::schema::ObTenantSchema *tenant_schema = NULL;
if (OB_ISNULL(GCTX.schema_service_)) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("unexpected error", KR(ret));
} else if (OB_FAIL(GCTX.schema_service_->get_tenant_schema_guard(
OB_SYS_TENANT_ID, schema_guard))) {
LOG_WARN("fail to get schema guard", KR(ret));
} else if (OB_FAIL(schema_guard.get_tenant_info(tenant_id_, tenant_schema))) {
LOG_WARN("failed to get tenant ids", KR(ret), K(tenant_id_));
} else if (OB_ISNULL(tenant_schema)) {
ret = OB_TENANT_NOT_EXIST;
LOG_WARN("tenant not exist", KR(ret), K(tenant_id_));
} else {
ObTenantLSInfo tenant_stat(GCTX.sql_proxy_, tenant_schema, tenant_id_,
GCTX.srv_rpc_proxy_, GCTX.lst_operator_);
if (OB_FAIL(tenant_stat.gather_stat(true))) {
LOG_WARN("failed to gather stat", KR(ret));
} else if (OB_FAIL(tenant_stat.create_new_ls_for_recovery(ls_attr.get_ls_id(),
ls_attr.get_ls_group_id(), ls_attr.get_create_scn(), trans))) {
LOG_WARN("failed to add new ls status info", KR(ret), K(ls_attr), K(sync_scn));
}
}
LOG_INFO("[LS_RECOVERY] create new ls", KR(ret), K(ls_attr));
}
return ret;
}
int ObRecoveryLSService::process_recovery_ls_manager()
{
int ret = OB_SUCCESS;
share::schema::ObSchemaGetterGuard schema_guard;
const share::schema::ObTenantSchema *tenant_schema = NULL;
if (OB_UNLIKELY(!inited_)) {
ret = OB_NOT_INIT;
LOG_WARN("not init", K(ret), K(inited_));
} else if (OB_ISNULL(GCTX.schema_service_)) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("unexpected error", KR(ret));
} else if (OB_FAIL(GCTX.schema_service_->get_tenant_schema_guard(
OB_SYS_TENANT_ID, schema_guard))) {
LOG_WARN("fail to get schema guard", KR(ret));
} else if (OB_FAIL(schema_guard.get_tenant_info(tenant_id_, tenant_schema))) {
LOG_WARN("failed to get tenant ids", KR(ret), K(tenant_id_));
} else if (OB_ISNULL(tenant_schema)) {
ret = OB_TENANT_NOT_EXIST;
LOG_WARN("tenant not exist", KR(ret), K(tenant_id_));
} else {
ObTenantLSInfo tenant_stat(GCTX.sql_proxy_, tenant_schema, tenant_id_,
GCTX.srv_rpc_proxy_, GCTX.lst_operator_);
if (OB_FAIL(tenant_stat.process_ls_stats_for_recovery())) {
LOG_WARN("failed to do process in standby", KR(ret), K(tenant_id_));
}
}
return ret;
}
int ObRecoveryLSService::check_can_do_recovery_(const ObAllTenantInfo &tenant_info)
{
int ret = OB_SUCCESS;
if (OB_UNLIKELY(!inited_)) {
ret = OB_NOT_INIT;
LOG_WARN("not init", K(ret), K(inited_));
} else if (OB_UNLIKELY(!tenant_info.is_valid())) {
ret = OB_INVALID_ARGUMENT;
LOG_WARN("tenant_info is invalid", KR(ret), K(tenant_info));
} else if (tenant_info.is_primary()) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("primary tenant can not do recovery", KR(ret), K(tenant_info));
} else if (tenant_info.is_standby()) {
//standby can do recovery
} else if (tenant_info.is_restore()) {
//need to check success to create init ls
share::ObPhysicalRestoreTableOperator restore_table_operator;
share::ObPhysicalRestoreJob job_info;
if (OB_ISNULL(proxy_)) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("sql can't null", K(ret), K(proxy_));
} else if (OB_FAIL(restore_table_operator.init(proxy_, tenant_id_))) {
LOG_WARN("fail to init restore table operator", KR(ret), K(tenant_id_));
} else if (OB_FAIL(restore_table_operator.get_job_by_tenant_id(tenant_id_,
job_info))) {
LOG_WARN("fail to get restore job", K(ret), K(tenant_id_));
} else if (job_info.is_valid_status_to_recovery()) {
//can do recovery
} else {
ret = OB_NEED_WAIT;
LOG_WARN("restore tenant not valid to recovery", KR(ret), K(job_info));
}
} else {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("unexpected tenant role", KR(ret), K(tenant_info));
}
return ret;
}
int ObRecoveryLSService::report_sys_ls_recovery_stat_(const SCN &sync_scn)
{
int ret = OB_SUCCESS;
if (OB_UNLIKELY(!inited_)) {
ret = OB_NOT_INIT;
LOG_WARN("not init", KR(ret), K(inited_));
} else if (OB_ISNULL(proxy_)) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("sql can't null", K(ret), K(proxy_));
} else if (SCN::base_scn() != sync_scn) {
ObLSRecoveryStatOperator ls_recovery;
ObLSRecoveryStat ls_recovery_stat;
if (OB_FAIL(construct_ls_recovery_stat(sync_scn, ls_recovery_stat))) {
LOG_WARN("failed to construct ls recovery stat", KR(ret), K(sync_scn));
} else if (OB_FAIL(ls_recovery.update_ls_recovery_stat(ls_recovery_stat,
*proxy_))) {
LOG_WARN("failed to update ls recovery stat", KR(ret),
K(ls_recovery_stat));
}
}
return ret;
}
}//end of namespace rootserver
}//end of namespace oceanbase