ORACLE的JOB調用存儲過程的邏輯方式的優勢是：快。但是單一的JOB執行某一項任務可能還是不能滿足性能需求。比如數據處理，可能在10幾分鐘需要處理的數據量是百萬級。下面介紹如何併發多子JOB執行分發數據。
PROCEDURE job_0125(a_job_counts IN INTEGER,
a_counts_per_job IN INTEGER,
a_timeout IN INTEGER) IS
e_business EXCEPTION;
v_job_name job_error_log.job_name%TYPE := 'job_0125';
v_sid NUMBER;
v_serial# NUMBER;
v_sqlcode job_error_log.error_no%TYPE;
v_sqlerrm job_error_log.error_message%TYPE;
v_error_comment job_error_log.error_comment%TYPE;
CURSOR v_job_running_cur IS
SELECT t.sid, t.serial#, t.job_id, t.create_time
FROM pscp_job_running_log t
WHERE t.job_status = '0'
AND t.job_name = 'job_0125'
AND t.create_time >= SYSDATE - 1;
v_job_running_record v_job_running_cur%ROWTYPE;
v_active_job_counts INTEGER := 0;
v_block_job_counts INTEGER := 0;
v_job_counts INTEGER;
v_group_id VARCHAR2(100);
v_group3_id VARCHAR2(50);
v_today DATE;
v_job_id NUMBER;
BEGIN
BEGIN
--獲取當前的SID和SERIAL#
SELECT sid, serial#
INTO v_sid, v_serial#
FROM v$session
WHERE sid = (SELECT MAX(sid)
FROM v$session
WHERE audsid = userenv('SESSIONID'));
EXCEPTION
WHEN NO_DATA_FOUND THEN
v_sqlcode := '-0000';
v_sqlerrm := 'JOB啓動異常：無法得到' || v_job_name || '的SID與SERIAL#';
v_error_comment := v_job_name || '啓動異常';
RAISE e_business;
END;
-- 檢查JOB輸入參數
IF a_job_counts < 1 OR a_job_counts > 20 THEN
v_sqlcode := '-0000';
v_sqlerrm := 'JOB啓動異常：JOB創建數量必須在1～20之間（包含1與20）';
v_error_comment := v_job_name || '啓動異常';
RAISE e_business;
END IF;
IF a_counts_per_job < 100 OR a_counts_per_job > 30000 THEN
v_sqlcode := '-0000';
v_sqlerrm := 'JOB啓動異常：每個JOB處理數據的數量必須在100～10000之間（包含100與10000）';
v_error_comment := v_job_name || '啓動異常';
RAISE e_business;
END IF;
IF a_timeout < 40 OR a_timeout > 480 THEN
v_sqlcode := '-0000';
v_sqlerrm := 'JOB啓動異常：JOB運行超時時間必須在60～480分鐘之間（包含60與480）';
v_error_comment := v_job_name || '啓動異常';
RAISE e_business;
END IF;
-- 分析當前JOB的運行情況
OPEN v_job_running_cur;
LOOP
FETCH v_job_running_cur
INTO v_job_running_record;
EXIT WHEN v_job_running_cur%NOTFOUND;
IF (SYSDATE - v_job_running_record.create_time) >=
a_timeout / (24 * 60) THEN
v_block_job_counts := v_block_job_counts + 1;
INSERT INTO job_error_log
(error_no,
error_message,
job_name,
job_user,
job_date,
error_comment,
sid,
serial#)
VALUES
('-0000',
'JOB運行異常：' || v_job_name || '創建的子JOB（' ||
v_job_running_record.job_id || '）運行時間超時',
v_job_name,
USER,
SYSDATE,
v_job_name || '創建的子JOB運行異常',
v_job_running_record.SID,
v_job_running_record.serial#);
ELSE
v_active_job_counts := v_active_job_counts + 1;
END IF;
END LOOP;
CLOSE v_job_running_cur;
-- 立即寫入JOB異常
IF v_block_job_counts > 0 THEN
COMMIT;
END IF;
-- 計算創建JOB的數量
v_job_counts := a_job_counts -
(v_block_job_counts + v_active_job_counts);
-- 判斷是否需要啓動新的子JOB
IF v_job_counts > 0 THEN
FOR v_index IN 1 .. v_job_counts LOOP
-- 爲JOB分配GROUP_ID，並更新處理狀態爲已分配
SELECT pscpdata.seq_group3_id.NEXTVAL INTO v_group3_id FROM DUAL;
v_group_id := '2011' || TO_CHAR(TRUNC(SYSDATE), 'YYYYMMDD') ||
v_group3_id;
v_today := SYSDATE;
UPDATE pscp_sms_wait_send t
SET t.disposal_status = '1', t.group_id = v_group_id
WHERE (t.disposal_status = '0' or t.disposal_status is null)
AND arrange_date <= SYSDATE + 1 / 24 / 60
AND arrange_date >= trunc(SYSDATE - 2)
AND ROWNUM <= a_counts_per_job;
IF SQL%FOUND THEN
INSERT INTO pscp_job_running_log
(group_id, username, create_time, job_status, job_name)
VALUES
(v_group_id, USER, v_today, '0', 'job_0125');
DBMS_JOB.submit(job => v_job_id,
what => 'PSCPCDE.pscp_sms_wait_to_sent_muti(' ||
v_group_id || ');',
next_date => SYSDATE,
INTERVAL => NULL);
COMMIT;
ELSE
ROLLBACK;
RETURN;
END IF;
END LOOP;
END IF;
EXCEPTION
WHEN OTHERS THEN
ROLLBACK;
IF v_sqlcode IS NULL THEN
v_sqlcode := SQLCODE;
v_sqlerrm := SUBSTR(SQLERRM, 1, 200);
END IF;
INSERT INTO job_error_log
(error_no,
error_message,
job_name,
job_user,
job_date,
error_comment,
sid,
serial#)
VALUES
(v_sqlcode,
v_sqlerrm,
v_job_name,
USER,
SYSDATE,
v_error_comment,
v_sid,
v_serial#);
COMMIT;
END job_0125;
END pkg_pscp_job_scheduler;CREATE OR REPLACE PROCEDURE pscp_sms_wait_to_sent_muti(a_group_id IN VARCHAR2) IS
counter NUMBER := 0; --統計數,當移動的數據達到10000條時,提交一次事務
sent_message_id integer; --發送表的message_id
v_today DATE;
v_sqlcode job_error_log.error_no%TYPE;
v_sqlerrm job_error_log.error_message%TYPE;
CURSOR v_job_info_cur IS
SELECT /*+rule*/
s.SID, s.serial#, s.logon_time, dj.job
FROM v$session s, dba_jobs_running djr, dba_jobs dj
WHERE djr.job = dj.job
AND s.SID = djr.SID
AND dj.what = 'PSCPCDE.PKG_PMCP_TO_PSCP_CONTROL.P_DATA_MANIPULATION(' ||
a_group_id || ');';
v_job_info_record v_job_info_cur%ROWTYPE;
BEGIN
-- 修改JOB狀態爲運行中
v_today := SYSDATE;
UPDATE pscp_job_running_log t
SET t.sid = v_job_info_record.sid,
t.serial# = v_job_info_record.serial#,
t.logon_time = v_job_info_record.logon_time,
t.job_id = v_job_info_record.job,
t.start_time = v_today,
t.end_time = v_today
WHERE t.group_id = a_group_id;
COMMIT;
--查詢本次的待發數據--
FOR cur IN (SELECT new_message_id, sms_type
FROM pscp_sms_wait_send t
WHERE t.group_id = a_group_id
AND t.disposal_status = '1'
AND arrange_date <= SYSDATE + 1 / 24 / 60
AND arrange_date >= trunc(SYSDATE - 2)) LOOP
BEGIN
select SEQ_MESSAGE_SENT_ID.NEXTVAL into sent_message_id from dual;
INSERT INTO pscp_sms_sent
(sent_message_id,
template_id,
version_id,
vsender_id,
send_user,
request_id,
mobile_number,
sms_content,
sms_priority,
arrive_date,
arrange_date,
expire_date,
mosequence_id,
src_id,
sender_series_id,
gateway_channel,
localgateway_date,
native_series_id,
template_name,
sms_task_id,
sms_task_num,
sms_benefit_serie,
confirm_template_id,
business_type_id,
party_no,
client_no,
group_task_no,
sms_operation_status,
url,
sms_type,
site_no,
linkid,
receive_msg_id,
TOTALNUMBER,
forbid_start_date,
forbid_stop_date,
stat_date)
(SELECT sent_message_id,
template_id,
version_id,
vsender_id,
send_user,
request_id,
mobile_number,
sms_content,
sms_priority,
trunc(SYSDATE),
arrange_date,
expire_date,
mosequence_id,
src_id,
sender_series_id,
gateway_channel,
localgateway_date,
native_series_id,
template_name,
sms_task_id,
sms_task_num,
sms_benefit_serie,
confirm_template_id,
business_type_id,
party_no,
client_no,
group_task_no,
sms_operation_status,
url,
sms_type,
site_no,
linkid,
receive_msg_id,
TOTALNUMBER,
forbid_start_date,
forbid_stop_date,
arrive_date AS stat_date
FROM pscp_sms_wait_send
WHERE new_message_id = cur.new_message_id);
DELETE FROM pscp_sms_wait_send
WHERE new_message_id = cur.new_message_id;
if cur.sms_type = 2 then
update pscp_tpl_mms_sent wt
set wt.mms_sent_id = sent_message_id
where mms_sent_id = cur.new_message_id;
end if;
--備份出現異常，給源表一個標識,繼續下一條得處理
EXCEPTION
WHEN OTHERS THEN
BEGIN
v_sqlcode := SQLCODE;
v_sqlerrm := SUBSTR(SQLERRM, 1, 200);
--插入違例表
INSERT INTO pscp_violate_history t
(t.violate_id,
t.template_name,
t.request_id,
t.violate_desp,
t.sms_content,
t.send_user,
t.sms_benefit_serie,
t.record_date,
t.mobile_number,
t.arrive_date,
t.expire_date,
t.sender_series_id,
t.result_type,
t.sms_send_template,
t.tpl_version_id,
t.sms_type,
t.vsender_id)
SELECT seq_violate_id.NEXTVAL,
TEMPLATE_NAME,
request_id,
v_sqlerrm,
sms_content,
send_user,
SMS_BENEFIT_SERIE,
SYSDATE,
mobile_number,
arrange_date,
expire_date,
sender_series_id,
3,
TEMPLATE_ID,
VERSION_ID,
sms_type,
VSENDER_ID
FROM pscp_sms_wait_send
WHERE new_message_id = cur.new_message_id;
DELETE FROM pscp_sms_wait_send pss
WHERE pss.new_message_id = cur.new_message_id;
EXCEPTION
WHEN OTHERS THEN
v_sqlerrm := SUBSTR(SQLERRM, 1, 200);
DELETE FROM pscp_sms_wait_send pss
WHERE pss.new_message_id = cur.new_message_id;
INSERT INTO pscp_db_log("prc_name", "log_desc", "log_time")
VALUES
('pscp_sms_wait_to_sent', v_sqlerrm, SYSDATE);
END;
END;
---1000條提交一次
counter := counter + 1;
if mod(counter, 10000) = 0 then
commit;
end if;
END LOOP;
-- 更新JOB狀態爲處理完成
UPDATE pscp_job_running_log t
SET t.end_time = SYSDATE, t.job_status = '1'
WHERE t.group_id = a_group_id;
COMMIT;
EXCEPTION
WHEN OTHERS THEN
v_sqlerrm := SUBSTR(SQLERRM, 1, 200);
INSERT INTO pscp_db_log("prc_name", "log_desc", "log_time")
VALUES
('pscp_sms_wait_to_sent', v_sqlerrm, SYSDATE);
COMMIT;
RAISE;
END pscp_sms_wait_to_sent_muti;