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

[SCN] fix farm failure

Browse Source
This commit is contained in:
obdev
2022-11-28 02:58:33 +00:00
committed by ob-robot
parent 87a9357186
commit 51de5b5911
594 changed files with 9722 additions and 7770 deletions
Download Patch File Download Diff File Expand all files Collapse all files

198
src/share/schema/ob_schema_service_sql_impl.cpp
View File

@ -886,7 +886,7 @@ int ObSchemaServiceSQLImpl::get_mock_fk_parent_table_schema_from_inner_table(
#define FETCH_ALL_TENANT_HISTORY_SQL3 "SELECT * from %s where 1 = 1"
#define FETCH_ALL_TABLE_HISTORY_SQL3 COMMON_SQL_WITH_TENANT
#define FETCH_ALL_TABLE_HISTORY_FULL_SCHEMA "SELECT /*+ leading(b a) use_nl(b a) %s*/ a.* FROM %s AS a JOIN "\
#define FETCH_ALL_TABLE_HISTORY_FULL_SCHEMA "SELECT /*+ leading(b a) use_nl(b a) no_rewrite() */ a.* FROM %s AS a JOIN "\
"(SELECT tenant_id, table_id, MAX(schema_version) AS schema_version FROM %s "\
"WHERE tenant_id = %lu AND schema_version <= %ld GROUP BY tenant_id, table_id) AS b "\
"ON a.tenant_id = b.tenant_id AND a.table_id = b.table_id AND a.schema_version = b.schema_version "\
@ -1119,49 +1119,18 @@ GET_ALL_SCHEMA_FUNC_DEFINE(sequence, ObSequenceSchema);
GET_ALL_SCHEMA_FUNC_DEFINE(dblink, ObDbLinkSchema);
GET_ALL_SCHEMA_FUNC_DEFINE(context, ObContextSchema);
GET_ALL_SCHEMA_FUNC_DEFINE(mock_fk_parent_table, ObSimpleMockFKParentTableSchema);
#define GET_ALL_SCHEMA_IGNORE_FAIL_FUNC_DEFINE(SCHEMA, SCHEMA_TYPE, CLUSTER_VERSION) \
int ObSchemaServiceSQLImpl::get_all_##SCHEMA##s(ObISQLClient &client, \
const ObRefreshSchemaStatus &schema_status, \
const int64_t schema_version, \
const uint64_t tenant_id, \
ObIArray<SCHEMA_TYPE> &schema_array) \
{ \
int ret = OB_SUCCESS; \
schema_array.reset(); \
if (!check_inner_stat()) { \
ret = OB_NOT_INIT; \
LOG_WARN("check inner stat fail"); \
} else if (!ObSchemaService::g_liboblog_mode_ \
&& GCONF.in_upgrade_mode() \
&& GET_MIN_CLUSTER_VERSION() < CLUSTER_VERSION) { \
/* system table in tenant space will be created in upgrade post stage after ver 2.2.1.*/ \
/* To avoid error while restart observer in upgrade stage, we skip refresh of new schema */ \
/* when cluster is still in upgrade stage. */ \
/* In addition, new schema history can be recorded after upgradation. */ \
LOG_INFO("ignore "#SCHEMA" schema table NOT EXIST error while upgrade", K(ret), K(schema_status), K(schema_version), K(tenant_id)); \
} else if (OB_FAIL(fetch_##SCHEMA##s(client, schema_status, schema_version, tenant_id, schema_array))) { \
LOG_WARN("fetch "#SCHEMA"s failed", K(ret), K(schema_status), K(schema_version), K(tenant_id)); \
/* for liboblog compatibility */ \
if (-ER_NO_SUCH_TABLE == ret && ObSchemaService::g_liboblog_mode_) { \
LOG_WARN("liboblog mode, ignore "#SCHEMA" schema table NOT EXIST error", K(ret), K(schema_status), K(schema_version), K(tenant_id)); \
ret = OB_SUCCESS; \
} \
} \
return ret; \
}
GET_ALL_SCHEMA_IGNORE_FAIL_FUNC_DEFINE(keystore, ObKeystoreSchema, CLUSTER_VERSION_2220);
GET_ALL_SCHEMA_IGNORE_FAIL_FUNC_DEFINE(tablespace, ObTablespaceSchema, CLUSTER_VERSION_2220);
GET_ALL_SCHEMA_IGNORE_FAIL_FUNC_DEFINE(trigger, ObSimpleTriggerSchema, CLUSTER_VERSION_2220);
GET_ALL_SCHEMA_IGNORE_FAIL_FUNC_DEFINE(label_se_policy, ObLabelSePolicySchema, CLUSTER_VERSION_2230);
GET_ALL_SCHEMA_IGNORE_FAIL_FUNC_DEFINE(label_se_component, ObLabelSeComponentSchema, CLUSTER_VERSION_2230);
GET_ALL_SCHEMA_IGNORE_FAIL_FUNC_DEFINE(label_se_label, ObLabelSeLabelSchema, CLUSTER_VERSION_2230);
GET_ALL_SCHEMA_IGNORE_FAIL_FUNC_DEFINE(label_se_user_level, ObLabelSeUserLevelSchema, CLUSTER_VERSION_2230);
GET_ALL_SCHEMA_IGNORE_FAIL_FUNC_DEFINE(profile, ObProfileSchema, CLUSTER_VERSION_2230);
GET_ALL_SCHEMA_IGNORE_FAIL_FUNC_DEFINE(audit, ObSAuditSchema, CLUSTER_VERSION_2230);
GET_ALL_SCHEMA_IGNORE_FAIL_FUNC_DEFINE(sys_priv, ObSysPriv, CLUSTER_VERSION_2250);
GET_ALL_SCHEMA_IGNORE_FAIL_FUNC_DEFINE(obj_priv, ObObjPriv, CLUSTER_VERSION_2260);
GET_ALL_SCHEMA_IGNORE_FAIL_FUNC_DEFINE(directory, ObDirectorySchema, CLUSTER_VERSION_4_0_0_0); // TODO: change version to 3200
GET_ALL_SCHEMA_FUNC_DEFINE(keystore, ObKeystoreSchema);
GET_ALL_SCHEMA_FUNC_DEFINE(tablespace, ObTablespaceSchema);
GET_ALL_SCHEMA_FUNC_DEFINE(trigger, ObSimpleTriggerSchema);
GET_ALL_SCHEMA_FUNC_DEFINE(label_se_policy, ObLabelSePolicySchema);
GET_ALL_SCHEMA_FUNC_DEFINE(label_se_component, ObLabelSeComponentSchema);
GET_ALL_SCHEMA_FUNC_DEFINE(label_se_label, ObLabelSeLabelSchema);
GET_ALL_SCHEMA_FUNC_DEFINE(label_se_user_level, ObLabelSeUserLevelSchema);
GET_ALL_SCHEMA_FUNC_DEFINE(profile, ObProfileSchema);
GET_ALL_SCHEMA_FUNC_DEFINE(audit, ObSAuditSchema);
GET_ALL_SCHEMA_FUNC_DEFINE(sys_priv, ObSysPriv);
GET_ALL_SCHEMA_FUNC_DEFINE(obj_priv, ObObjPriv);
GET_ALL_SCHEMA_FUNC_DEFINE(directory, ObDirectorySchema);
int ObSchemaServiceSQLImpl::get_all_db_privs(ObISQLClient &client,
const ObRefreshSchemaStatus &schema_status,
@ -3730,52 +3699,46 @@ int ObSchemaServiceSQLImpl::fetch_all_udt_object_info(
ObMySQLResult *result = NULL;
ObSqlString sql;
const uint64_t exec_tenant_id = fill_exec_tenant_id(schema_status);
if ((ObSchemaService::g_liboblog_mode_ && GET_MIN_CLUSTER_VERSION() < CLUSTER_VERSION_2270)
|| (GCONF.in_upgrade_mode() && GET_MIN_CLUSTER_VERSION() < CLUSTER_VERSION_2270)) {
LOG_WARN("in upgrade mode, table maybe not ready, just skip",
K(ret), K(schema_status), K(tenant_id), K(schema_version));
} else {
if (OB_FAIL(sql.append_fmt(FETCH_ALL_OBJECT_TYPE_HISTORY_SQL,
OB_ALL_TENANT_OBJECT_TYPE_HISTORY_TNAME,
fill_extract_tenant_id(schema_status, tenant_id)))) {
if (OB_FAIL(sql.append_fmt(FETCH_ALL_OBJECT_TYPE_HISTORY_SQL,
OB_ALL_TENANT_OBJECT_TYPE_HISTORY_TNAME,
fill_extract_tenant_id(schema_status, tenant_id)))) {
LOG_WARN("append sql failed", K(ret));
} else if (NULL != object_ids && object_ids_size > 0) {
if (OB_FAIL(sql.append_fmt(" AND OBJECT_TYPE_ID IN ("))) {
LOG_WARN("append sql failed", K(ret));
} else if (NULL != object_ids && object_ids_size > 0) {
if (OB_FAIL(sql.append_fmt(" AND OBJECT_TYPE_ID IN ("))) {
LOG_WARN("append sql failed", K(ret));
} else {
for (int64_t i = 0; OB_SUCC(ret) && i < object_ids_size; ++i) {
const uint64_t object_type_id = fill_extract_schema_id(schema_status, object_ids[i]);
if (OB_FAIL(sql.append_fmt("%s%lu", 0 == i ? "" : ", ", object_type_id))) {
LOG_WARN("append sql failed", K(ret), K(i));
}
}
if (OB_SUCC(ret)) {
if (OB_FAIL(sql.append(")"))) {
LOG_WARN("append sql failed", K(ret));
}
} else {
for (int64_t i = 0; OB_SUCC(ret) && i < object_ids_size; ++i) {
const uint64_t object_type_id = fill_extract_schema_id(schema_status, object_ids[i]);
if (OB_FAIL(sql.append_fmt("%s%lu", 0 == i ? "" : ", ", object_type_id))) {
LOG_WARN("append sql failed", K(ret), K(i));
}
}
} else { }
if (OB_SUCC(ret)) {
if (OB_FAIL(sql.append_fmt(" AND SCHEMA_VERSION <= %ld", schema_version))) {
LOG_WARN("append failed", K(ret));
} else if (OB_FAIL(
sql.append(" ORDER BY TENANT_ID ASC, OBJECT_TYPE_ID ASC, TYPE ASC, SCHEMA_VERSION DESC"))) {
LOG_WARN("sql append failed", K(ret));
if (OB_SUCC(ret)) {
if (OB_FAIL(sql.append(")"))) {
LOG_WARN("append sql failed", K(ret));
}
}
}
if (OB_SUCC(ret)) {
const int64_t snapshot_timestamp = schema_status.snapshot_timestamp_;
DEFINE_SQL_CLIENT_RETRY_WEAK_WITH_SNAPSHOT(sql_client, snapshot_timestamp);
if (OB_FAIL(sql_client_retry_weak.read(res, exec_tenant_id, sql.ptr()))) {
LOG_WARN("execute sql failed", K(ret), K(tenant_id), K(sql));
} else if (OB_UNLIKELY(NULL == (result = res.get_result()))) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("Fail to get result", K(ret));
} else if (OB_FAIL(ObSchemaRetrieveUtils::retrieve_udt_object_schema(tenant_id,
*result, udt_infos))) {
LOG_WARN("Failed to retrieve udt object type infos", K(ret));
}
} else { }
if (OB_SUCC(ret)) {
if (OB_FAIL(sql.append_fmt(" AND SCHEMA_VERSION <= %ld", schema_version))) {
LOG_WARN("append failed", K(ret));
} else if (OB_FAIL(
sql.append(" ORDER BY TENANT_ID ASC, OBJECT_TYPE_ID ASC, TYPE ASC, SCHEMA_VERSION DESC"))) {
LOG_WARN("sql append failed", K(ret));
}
}
if (OB_SUCC(ret)) {
const int64_t snapshot_timestamp = schema_status.snapshot_timestamp_;
DEFINE_SQL_CLIENT_RETRY_WEAK_WITH_SNAPSHOT(sql_client, snapshot_timestamp);
if (OB_FAIL(sql_client_retry_weak.read(res, exec_tenant_id, sql.ptr()))) {
LOG_WARN("execute sql failed", K(ret), K(tenant_id), K(sql));
} else if (OB_UNLIKELY(NULL == (result = res.get_result()))) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("Fail to get result", K(ret));
} else if (OB_FAIL(ObSchemaRetrieveUtils::retrieve_udt_object_schema(tenant_id,
*result, udt_infos))) {
LOG_WARN("Failed to retrieve udt object type infos", K(ret));
}
}
}
@ -3903,12 +3866,7 @@ int ObSchemaServiceSQLImpl::fetch_all_user_info(
}
}
if (OB_SUCC(ret)) {
if (!ObSchemaService::g_liboblog_mode_
&& GCONF.in_upgrade_mode()
&& GET_MIN_CLUSTER_VERSION() < CLUSTER_VERSION_2260) {
LOG_WARN("in upgrade mode, table maybe not ready, just skip",
K(ret), K(schema_status), K(tenant_id), K(schema_version));
} else if (OB_FAIL(fetch_role_grantee_map_info(schema_status,
if (OB_FAIL(fetch_role_grantee_map_info(schema_status,
schema_version,
tenant_id,
sql_client,
@ -4270,15 +4228,8 @@ int ObSchemaServiceSQLImpl::fetch_tables(
schema_service_))) {
LOG_WARN("fail to get all table name", K(ret), K(exec_tenant_id));
} else if (!is_increase_schema) {
/* no_rewrite() hint for the following sql stmt is needed since ver 2.2.1.
* Otherwise, the following sql stmt will run in low performance.
* To avoid possible problem (no_rewrite() hint may cause error in ver 1.4.21),
* no_rewrite() hint is only added since ver 2.2.1.
*/
const char *addition_hint = GET_MIN_CLUSTER_VERSION() < CLUSTER_VERSION_2210 ?
"" : " no_rewrite() ";
if (OB_FAIL(sql.append_fmt(FETCH_ALL_TABLE_HISTORY_FULL_SCHEMA,
addition_hint, table_name, table_name,
table_name, table_name,
fill_extract_tenant_id(schema_status, tenant_id),
schema_version,
fill_extract_schema_id(schema_status, OB_ALL_CORE_TABLE_TID)))) {
@ -4940,11 +4891,7 @@ int ObSchemaServiceSQLImpl::get_not_core_table_schema(
}
}
if (OB_SUCCESS == ret) {
if (!ObSchemaService::g_liboblog_mode_ &&
GCONF.in_upgrade_mode() &&
GET_MIN_CLUSTER_VERSION() < CLUSTER_VERSION_2220) {
LOG_WARN("skip fetch trigger list while upgrade in process from 221 to 222", K(tenant_id));
} else if (OB_FAIL(fetch_trigger_list(schema_status, tenant_id, table_id,
if (OB_FAIL(fetch_trigger_list(schema_status, tenant_id, table_id,
schema_version, sql_client, *table_schema))) {
LOG_WARN("Failed to fetch trigger list", K(ret));
if (-ER_NO_SUCH_TABLE == ret && ObSchemaService::g_liboblog_mode_) {
@ -5126,13 +5073,7 @@ int ObSchemaServiceSQLImpl::fetch_constraint_column_info(const ObRefreshSchemaSt
LOG_WARN("append table_id and foreign key id and schema version",
K(ret), K(table_id), K(cst->get_constraint_id()), K(schema_version));
} else if (OB_FAIL(sql_client_retry_weak.read(res, exec_tenant_id, sql.ptr()))) {
if ((OB_TABLE_NOT_EXIST == ret) && (GET_MIN_CLUSTER_VERSION() < CLUSTER_VERSION_3100)) {
LOG_INFO("cluster version is earlier than 3.1.0, ignore OB_TABLE_NOT_EXIST",
K(ret), K(GET_MIN_CLUSTER_VERSION()), K(sql));
ret = OB_SUCCESS;
} else {
LOG_WARN("execute sql failed", K(ret), K(GET_MIN_CLUSTER_VERSION()), K(sql));
}
LOG_WARN("execute sql failed", K(ret), K(sql));
} else if (OB_ISNULL(result = res.get_result())) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("failed to get result. ", K(ret));
@ -5327,13 +5268,7 @@ int ObSchemaServiceSQLImpl::fetch_part_info(
const bool check_deleted = true;
DEFINE_SQL_CLIENT_RETRY_WEAK_WITH_SNAPSHOT(sql_client, snapshot_timestamp);
if (OB_FAIL(sql_client_retry_weak.read(res, exec_tenant_id, sql.ptr()))) {
if ((OB_TABLE_NOT_EXIST == ret) && (GET_MIN_CLUSTER_VERSION() < CLUSTER_VERSION_3100)) {
LOG_INFO("cluster version is earlier than 3.1.0, ignore OB_TABLE_NOT_EXIST",
K(ret), K(GET_MIN_CLUSTER_VERSION()), K(sql));
ret = OB_SUCCESS;
} else {
LOG_WARN("execute sql failed", K(ret), K(GET_MIN_CLUSTER_VERSION()), K(sql));
}
LOG_WARN("execute sql failed", K(ret), K(sql));
} else if (OB_UNLIKELY(NULL == (result = res.get_result()))) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("fail to get result. ", K(sql), K(ret));
@ -5884,18 +5819,9 @@ int ObSchemaServiceSQLImpl::fetch_foreign_key_column_info(
if (OB_FAIL(sql.append_fmt(FETCH_ALL_FOREIGN_KEY_COLUMN_HISTORY_SQL, OB_ALL_FOREIGN_KEY_COLUMN_HISTORY_TNAME,
fill_extract_tenant_id(schema_status, tenant_id)))) {
LOG_WARN("append sql failed", K(ret));
} else if ((!ObSchemaService::g_liboblog_mode_
|| (ObSchemaService::g_liboblog_mode_ && GET_MIN_CLUSTER_VERSION() >= CLUSTER_VERSION_2250))
&& OB_FAIL(sql.append_fmt(" AND foreign_key_id = %lu AND schema_version <= %ld ORDER BY position ASC, schema_version DESC",
fill_extract_schema_id(schema_status, foreign_key_info.foreign_key_id_),
schema_version))) {
LOG_WARN("append table_id and foreign key id and schema version",
K(ret), K(foreign_key_info.foreign_key_id_), K(schema_version));
} else if (ObSchemaService::g_liboblog_mode_
&& GET_MIN_CLUSTER_VERSION() < CLUSTER_VERSION_2250
&& OB_FAIL(sql.append_fmt(" AND foreign_key_id = %lu AND schema_version <= %ld",
fill_extract_schema_id(schema_status, foreign_key_info.foreign_key_id_),
schema_version))) {
} else if (OB_FAIL(sql.append_fmt(" AND foreign_key_id = %lu AND schema_version <= %ld ORDER BY position ASC, schema_version DESC",
fill_extract_schema_id(schema_status, foreign_key_info.foreign_key_id_),
schema_version))) {
LOG_WARN("append table_id and foreign key id and schema version",
K(ret), K(foreign_key_info.foreign_key_id_), K(schema_version));
} else if (OB_FAIL(sql_client_retry_weak.read(res, exec_tenant_id, sql.ptr()))) {
@ -5928,10 +5854,7 @@ int ObSchemaServiceSQLImpl::fetch_foreign_key_array_for_simple_table_schemas(
DEFINE_SQL_CLIENT_RETRY_WEAK_WITH_SNAPSHOT(sql_client, snapshot_timestamp);
const uint64_t exec_tenant_id = fill_exec_tenant_id(schema_status);
// FIXME: The following SQL will cause full table scan, which it's poor performance when the amount of table data is large.
if (ObSchemaService::g_liboblog_mode_ && GET_MIN_CLUSTER_VERSION() < CLUSTER_VERSION_2100) {
LOG_INFO("liboblog mode, ignore");
ret = OB_SUCCESS;
} else if (OB_FAIL(sql.append_fmt(FETCH_TABLE_ID_AND_NAME_FROM_ALL_FOREIGN_KEY_SQL, OB_ALL_FOREIGN_KEY_HISTORY_TNAME,
if (OB_FAIL(sql.append_fmt(FETCH_TABLE_ID_AND_NAME_FROM_ALL_FOREIGN_KEY_SQL, OB_ALL_FOREIGN_KEY_HISTORY_TNAME,
fill_extract_tenant_id(schema_status, tenant_id)))) {
LOG_WARN("failed to append sql", K(ret), K(tenant_id));
} else if (OB_FAIL(sql.append_fmt(" AND child_table_id IN "))) {
@ -6081,10 +6004,7 @@ int ObSchemaServiceSQLImpl::fetch_constraint_array_for_simple_table_schemas(cons
SMART_VAR(ObMySQLProxy::MySQLResult, res) {
ObMySQLResult *result = NULL;
DEFINE_SQL_CLIENT_RETRY_WEAK_WITH_SNAPSHOT(sql_client, snapshot_timestamp);
if (ObSchemaService::g_liboblog_mode_ && GET_MIN_CLUSTER_VERSION() < CLUSTER_VERSION_2110) {
LOG_INFO("liboblog mode, ignore");
ret = OB_SUCCESS;
} else if (OB_FAIL(sql.append_fmt(FETCH_TABLE_ID_AND_CST_NAME_FROM_ALL_CONSTRAINT_HISTORY_SQL, OB_ALL_CONSTRAINT_HISTORY_TNAME,
if (OB_FAIL(sql.append_fmt(FETCH_TABLE_ID_AND_CST_NAME_FROM_ALL_CONSTRAINT_HISTORY_SQL, OB_ALL_CONSTRAINT_HISTORY_TNAME,
fill_extract_tenant_id(schema_status, tenant_id)))) {
LOG_WARN("failed to append sql", K(ret), K(tenant_id));
} else if (OB_FAIL(sql.append_fmt(" AND table_id IN "))) {
@ -8058,7 +7978,7 @@ int ObSchemaServiceSQLImpl::fetch_link_table_info(const ObDbLinkSchema &dblink_s
int16_t scale = 0;
int32_t length = 0;
ObString column_name;
bool old_max_length = (GET_MIN_CLUSTER_VERSION() < CLUSTER_VERSION_1470) && !ObSchemaService::g_liboblog_mode_;
bool old_max_length = false;
if (OB_FAIL(result->get_col_meta(i, old_max_length, column_name, type, precision, scale, length))) {
LOG_WARN("failed to get column meta", K(i), K(old_max_length), K(ret));
} else if (OB_FAIL(column_schema.set_column_name(column_name))) {