「事件驅動架構」GoldenGate創建從Oracle到Kafka的CDC事件流(2)

前文：「事件驅動架構」GoldenGate創建從Oracle到Kafka的CDC事件流(1)

步驟7/12:安裝並運行Apache Kafka

從VM的桌面環境中打開Firefox並下載Apache Kafka(我使用的是kafka_2.11-2.1.1.tgz)。

現在，打開一個Linux shell並重置CLASSPATH環境變量(在BigDataLite-4.11虛擬機中設置的當前值會在Kafka中產生衝突):

declare -x CLASSPATH=""

從同一個Linux shell中，解壓縮壓縮包，啟動ZooKeeper和Kafka:

cdtar zxvf Downloads/kafka_2.11-2.1.1.tgzcd kafka_2.11-2.1.1./bin/zookeeper-server-start.sh -daemon config/zookeeper.properties./bin/kafka-server-start.sh -daemon config/server.properties

你可以通過啟動「echo stats | nc localhost 2181」來檢查ZooKeeper是否正常:

[oracle@bigdatalite ~]$ echo stats | nc localhost 2181Zookeeper version: 3.4.5-cdh5.13.1--1, built on 11/09/2017 16:28 GMTClients: /127.0.0.1:34997[1](queued=0,recved=7663,sent=7664) /0:0:0:0:0:0:0:1:17701[0](queued=0,recved=1,sent=0) Latency min/avg/max: 0/0/25Received: 8186Sent: 8194Connections: 2Outstanding: 0Zxid: 0x3fMode: standaloneNode count: 25

您可以檢查Kafka是否與「echo dump | nc localhost 2181 | grep代理」(一個字符串/brokers/ids/0應該出現)

[oracle@bigdatalite ~]$ echo dump | nc localhost 2181 | grep brokers/brokers/ids/0

用於PoC的BigDataLite-4.11虛擬機已經在啟動虛擬機時啟動了一個較老的ZooKeeper實例。因此，請確保禁用了步驟1中描述的所有服務。此外，當您打開一個新的Linux shell時，請注意在啟動ZooKeeper和Kafka之前總是要重置CLASSPATH環境變量，這一點在步驟開始時已經解釋過了。

步驟8/12:為大數據安裝GoldenGate

同樣，從這個頁面下載Oracle GoldenGate for Big Data 12c只需要使用VM中安裝的Firefox瀏覽器(我在Linux x86-64上使用Oracle GoldenGate for Big Data 12.3.2.1.1)。請注意，您需要一個(免費)Oracle帳戶來獲得它。

安裝很容易，只是爆炸壓縮包內的下載:

cd ~/Downloadsunzip OGG_BigData_Linux_x64_12.3.2.1.1.zipcd ..mkdir ogg-bd-poccd ogg-bd-poctar xvf ../Downloads/OGG_BigData_Linux_x64_12.3.2.1.1.tar

就這樣，GoldenGate for Big Data 12c被安裝在/home/oracle/ogg-bd-poc文件夾中。

同樣，BigDataLite-4.11虛擬機已經在/u01/ogg-bd文件夾中安裝了用於大數據的GoldenGate。但它是一個較舊的版本，連接Kafka的選項較少。

步驟9/12:啟動GoldenGate for Big Data Manager

打開大數據大門

cd ~/ogg-bd-poc./ggsci

需要更改管理器埠，否則之前啟動的與GoldenGate (classic)管理器的衝突將被引發。

因此，從大數據的GoldenGate來看，CLI運行:

create subdirsedit params mgr

一個vi實例將開始，只是寫這個內容:

PORT 27801

然後保存內容，退出vi，返回CLI，我們終於可以啟動GoldenGate for Big Data manager監聽埠27081:

步驟10/12:創建數據泵(Data Pump)

現在，我們需要創建在GoldenGate世界中被稱為數據泵的東西。數據泵是一個提取過程，它監視一個跟蹤日誌，並(實時地)將任何更改推到另一個由不同的(通常是遠程的)GoldenGate實例管理的跟蹤日誌。

對於這個PoC，由GoldenGate (classic)管理的trail log aa將被泵送至GoldenGate管理的trail log bb進行大數據處理。

因此，如果您關閉它，請回到來自Linux shell的GoldenGate(經典)CLI:

cd /u01/ogg./ggsci

來自GoldenGate(經典)CLI:

edit params pmpeshop

並在vi中加入以下內容:

EXTRACT pmpeshopUSERIDALIAS ggadminSETENV (ORACLE_SID='orcl')-- GoldenGate for Big Data address/port:RMTHOST localhost, MGRPORT 27801RMTTRAIL ./dirdat/bbPASSTHRU-- The "tokens" part it is useful for writing in the Kafka messages-- the Transaction ID and the database Change Serial NumberTABLE orcl.eshop.*, tokens(txid = @GETENV('TRANSACTION', 'XID'), csn = @GETENV('TRANSACTION', 'CSN'));

保存內容並退出vi。

正如已經解釋的提取器，保存的內容將存儲在/u01/ogg/dirprm/pmpeshop中。人口、難民和移民事務局文件。

現在我們要註冊並啟動數據泵，從GoldenGate CLI:

dblogin useridalias ggadminadd extract pmpeshop, exttrailsource ./dirdat/aa begin nowadd rmttrail ./dirdat/bb extract pmpeshopstart pmpeshop

通過從CLI運行以下命令之一來檢查數據泵的狀態:

info pmpeshopview report pmpeshop

你甚至可以在金門大數據的dirdat文件夾中查看trail log bb是否已經創建:

[oracle@bigdatalite dirdat]$ ls -l ~/ogg-bd-poc/dirdattotal 0-rw-r. 1 oracle oinstall 0 May 30 13:22 bb000000000[oracle@bigdatalite dirdat]$

那檢查泵送過程呢?來自Linux shell:

sqlplus eshop/eshop@ORCL

執行這個SQL腳本創建一個新的模擬客戶訂單:

INSERT INTO CUSTOMER_ORDER (ID, CODE, CREATED, STATUS, UPDATE_TIME)VALUES (CUSTOMER_ORDER_SEQ.NEXTVAL, 'AAAA02', SYSDATE, 'SHIPPING', SYSTIMESTAMP); INSERT INTO CUSTOMER_ORDER_ITEM (ID, ID_CUSTOMER_ORDER, DESCRIPTION, QUANTITY)VALUES (CUSTOMER_ORDER_ITEM_SEQ.NEXTVAL, CUSTOMER_ORDER_SEQ.CURRVAL, 'Inside Out', 1); COMMIT;

現在從GoldenGate(經典)CLI運行:

stats pmpeshop

用於檢查插入操作是否正確計數(在輸出的一部分下面):

GGSCI (bigdatalite.localdomain as ggadmin@cdb/CDB$ROOT) 11> stats pmpeshop Sending STATS request to EXTRACT PMPESHOP ... Start of Statistics at 2019-05-30 14:49:00. Output to ./dirdat/bb: Extracting from ORCL.ESHOP.CUSTOMER_ORDER to ORCL.ESHOP.CUSTOMER_ORDER: *** Total statistics since 2019-05-30 14:01:56 ***Total inserts 1.00Total updates 0.00Total deletes 0.00Total discards 0.00Total operations 1.00

此外，您還可以驗證GoldenGate中存儲的用於測試泵過程的大數據的跟蹤日誌的時間戳。事務提交後，從Linux shell運行:「ln -l ~/og -bd-poc/dirdat」，並檢查最後一個以「bb」作為前綴的文件的時間戳。

步驟11/12:將事務發布到Kafka

最後，我們將在GoldenGate中為BigData創建一個副本流程，以便在Kafka主題中發布泵出的業務事務。replicat將從trail日誌bb讀取事務中的插入、更新和刪除操作，並將它們轉換為JSON編碼的Kafka消息。

因此，創建一個名為eshop_kafkaconnect的文件。文件夾/home/oracle/ogg-bd- pocd /dirprm中的屬性包含以下內容:

# File: /home/oracle/ogg-bd-poc/dirprm/eshop_kafkaconnect.properties# ---- # address/port of the Kafka brokerbootstrap.servers=localhost:9092acks=1 #JSON Converter Settingskey.converter=org.apache.kafka.connect.json.JsonConverterkey.converter.schemas.enable=falsevalue.converter=org.apache.kafka.connect.json.JsonConvertervalue.converter.schemas.enable=false #Adjust for performancebuffer.memory=33554432batch.size=16384linger.ms=0 # This property fix a start-up error as explained by Oracle Support here:# https://support.oracle.com/knowledge/Middleware/2455697_1.htmlconverter.type=key

在同一個文件夾中，創建一個名為eshop_kc的文件。具有以下內容的道具:

# File: /home/oracle/ogg-bd-poc/dirprm/eshop_kc.props# -gg.handlerlist=kafkaconnect #The handler propertiesgg.handler.kafkaconnect.type=kafkaconnectgg.handler.kafkaconnect.kafkaProducerConfigFile=eshop_kafkaconnect.propertiesgg.handler.kafkaconnect.mode=tx #The following selects the topic name based only on the schema namegg.handler.kafkaconnect.topicMappingTemplate=CDC-${schemaName} #The following selects the message key using the concatenated primary keysgg.handler.kafkaconnect.keyMappingTemplate=${primaryKeys} #The formatter propertiesgg.handler.kafkaconnect.messageFormatting=opgg.handler.kafkaconnect.insertOpKey=Igg.handler.kafkaconnect.updateOpKey=Ugg.handler.kafkaconnect.deleteOpKey=Dgg.handler.kafkaconnect.truncateOpKey=Tgg.handler.kafkaconnect.treatAllColumnsAsStrings=falsegg.handler.kafkaconnect.iso8601Format=falsegg.handler.kafkaconnect.pkUpdateHandling=abendgg.handler.kafkaconnect.includeTableName=truegg.handler.kafkaconnect.includeOpType=truegg.handler.kafkaconnect.includeOpTimestamp=truegg.handler.kafkaconnect.includeCurrentTimestamp=truegg.handler.kafkaconnect.includePosition=truegg.handler.kafkaconnect.includePrimaryKeys=truegg.handler.kafkaconnect.includeTokens=true goldengate.userexit.writers=javawriterjavawriter.stats.display=TRUEjavawriter.stats.full=TRUE gg.log=log4jgg.log.level=INFO gg.report.time=30sec # Apache Kafka Classpath# Put the path of the "libs" folder inside the Kafka home pathgg.classpath=/home/oracle/kafka_2.11-2.1.1/libs/* javawriter.bootoptions=-Xmx512m -Xms32m -Djava.class.path=.:ggjava/ggjava.jar:./dirprm

如果關閉，重啟大數據CLI的GoldenGate:

cd ~/ogg-bd-poc./ggsci

and start to create a replicat from the CLI with:

edit params repeshop

in viput this content:

REPLICAT repeshopTARGETDB LIBFILE libggjava.so SET property=dirprm/eshop_kc.propsGROUPTRANSOPS 1000MAP orcl.eshop.*, TARGET orcl.eshop.*;

然後保存內容並退出vi。現在將replicat與trail log bb關聯，並使用以下命令啟動replicat進程，以便從GoldenGate啟動大數據CLI:

add replicat repeshop, exttrail ./dirdat/bbstart repeshop

Check that the replicatis live and kicking with one of these commands:

info repeshopview report repeshop

Now, connect to the ESHOP schema from another Linux shell:

sqlplus eshop/eshop@ORCL

and commit something:

INSERT INTO CUSTOMER_ORDER (ID, CODE, CREATED, STATUS, UPDATE_TIME)VALUES (CUSTOMER_ORDER_SEQ.NEXTVAL, 'AAAA03', SYSDATE, 'DELIVERED', SYSTIMESTAMP); INSERT INTO CUSTOMER_ORDER_ITEM (ID, ID_CUSTOMER_ORDER, DESCRIPTION, QUANTITY)VALUES (CUSTOMER_ORDER_ITEM_SEQ.NEXTVAL, CUSTOMER_ORDER_SEQ.CURRVAL, 'Cars 3', 2); COMMIT;

From the GoldenGate for Big Data CLI, check that the INSERT operation was counted for the replicatprocess by running:

stats repeshop

And (hurrah!) we can have a look inside Kafka, as the Linux shell checks that the topic named CDC-ESHOPwas created:

cd ~/kafka_2.11-2.1.1/bin./kafka-topics.sh --list --zookeeper localhost:2181

and from the same folder run the following command for showing the CDC events stored in the topic:

./kafka-console-consumer.sh --bootstrap-server localhost:9092 --topic CDC-ESHOP --from-beginning

You should see something like:

[oracle@bigdatalite kafka_2.11-2.1.1]$ ./bin/kafka-console-consumer.sh --bootstrap-server localhost:9092 --topic CDC-ESHOP --from-beginning {"table":"ORCL.ESHOP.CUSTOMER_ORDER","op_type":"I","op_ts":"2019-05-31 04:24:34.000327","current_ts":"2019-05-31 04:24:39.637000","pos":"00000000020000003830","primary_keys":["ID"],"tokens":{"txid":"9.32.6726","csn":"13906131"},"before":null,"after":{"ID":11.0,"CODE":"AAAA03","CREATED":"2019-05-31 04:24:34","STATUS":"DELIVERED","UPDATE_TIME":"2019-05-31 04:24:34.929950000"}}{"table":"ORCL.ESHOP.CUSTOMER_ORDER_ITEM","op_type":"I","op_ts":"2019-05-31 04:24:34.000327","current_ts":"2019-05-31 04:24:39.650000","pos":"00000000020000004074","primary_keys":["ID"],"tokens":{"txid":"9.32.6726","csn":"13906131"},"before":null,"after":{"ID":11.0,"ID_CUSTOMER_ORDER":11.0,"DESCRIPTION":"Cars 3","QUANTITY":2}}

For a better output, install jq:

sudo yum -y install jq./kafka-console-consumer.sh --bootstrap-server localhost:9092 --topic CDC-ESHOP --from-beginning | jq .

and here is how will appear the JSON events:

{ "table": "ORCL.ESHOP.CUSTOMER_ORDER", "op_type": "I", "op_ts": "2019-05-31 04:24:34.000327", "current_ts": "2019-05-31 04:24:39.637000", "pos": "00000000020000003830", "primary_keys": [ "ID" ], "tokens": { "txid": "9.32.6726", "csn": "13906131" }, "before": null, "after": { "ID": 11, "CODE": "AAAA03", "CREATED": "2019-05-31 04:24:34", "STATUS": "DELIVERED", "UPDATE_TIME": "2019-05-31 04:24:34.929950000" }}{ "table": "ORCL.ESHOP.CUSTOMER_ORDER_ITEM", "op_type": "I", "op_ts": "2019-05-31 04:24:34.000327", "current_ts": "2019-05-31 04:24:39.650000", "pos": "00000000020000004074", "primary_keys": [ "ID" ], "tokens": { "txid": "9.32.6726", "csn": "13906131" }, "before": null, "after": { "ID": 11, "ID_CUSTOMER_ORDER": 11, "DESCRIPTION": "Cars 3", "QUANTITY": 2 }}

現在打開Kafka -console-consumer.sh進程，並在ESHOP上執行其他一些資料庫事務，以便實時列印發送給Kafka的CDC事件流。

以下是一些用於更新和刪除操作的JSON事件示例:

// Generated with: UPDATE CUSTOMER_ORDER SET STATUS='DELIVERED' WHERE ID=8; { "table": "ORCL.ESHOP.CUSTOMER_ORDER", "op_type": "U", "op_ts": "2019-05-31 06:22:07.000245", "current_ts": "2019-05-31 06:22:11.233000", "pos": "00000000020000004234", "primary_keys": [ "ID" ], "tokens": { "txid": "14.6.2656", "csn": "13913689" }, "before": { "ID": 8, "CODE": null, "CREATED": null, "STATUS": "SHIPPING", "UPDATE_TIME": null }, "after": { "ID": 8, "CODE": null, "CREATED": null, "STATUS": "DELIVERED", "UPDATE_TIME": null }} // Generated with: DELETE CUSTOMER_ORDER_ITEM WHERE ID=3;{ "table": "ORCL.ESHOP.CUSTOMER_ORDER_ITEM", "op_type": "D", "op_ts": "2019-05-31 06:25:59.000916", "current_ts": "2019-05-31 06:26:04.910000", "pos": "00000000020000004432", "primary_keys": [ "ID" ], "tokens": { "txid": "14.24.2651", "csn": "13913846" }, "before": { "ID": 3, "ID_CUSTOMER_ORDER": 1, "DESCRIPTION": "Toy Story", "QUANTITY": 1 }, "after": null}

恭喜你!你完成了PoC:

步驟12/12:使用PoC

GoldenGate中提供的Kafka Connect處理程序有很多有用的選項，可以根據需要定製集成。點擊這裡查看官方文件。

例如，您可以選擇為CDC流中涉及的每個表創建不同的主題，只需在eshop_kc.props中編輯此屬性:

gg.handler.kafkaconnect.topicMappingTemplate=CDC-${schemaName}-${tableName}

更改後重新啟動replicat，從GoldenGate for Big Data CLI:

stop repeshopstart repeshop

您可以在「~/og -bd-poc/AdapterExamples/big-data/kafka_connect」文件夾中找到其他配置示例。

結論

在本文中，我們通過GoldenGate技術在Oracle資料庫和Kafka代理之間創建了一個完整的集成。CDC事件流以Kafka實時發布。

為了簡單起見，我們使用了一個已經全部安裝的虛擬機，但是您可以在不同的主機上免費安裝用於大數據的GoldenGate和Kafka。

請在評論中告訴我您對這種集成的潛力(或限制)的看法。