Hive Transaction 事务性 小试

2017-01-05 11:11:58来源:oschina作者:王建奎Jerrick人点击

提到Hive一般都会想到,Hive是数据仓库,支持类SQL查询,有很多语法支持,可以嵌套MR,写Transform、写UDF/UDAF等,但是,不支持更新操作。所以Hive的常见也一般都是一次写入,频繁读取。从Hive 0.13开始,加入了ACID的新feature,但是0.13的时候还不支持insert、update和delete操作,我也并没有欣然的当小白鼠。


目前我们平台使用hive1.2.1的社区版,业务上也遇到了需要更新的场景。也是继续调研Transaction的特性。Transaction有几个依赖条件:


1、只支持ORCFile


2、默认关闭,需要自己开启


3、表必须分桶


4、0.14开始支持insert、update和delete


5、必须加入一大堆配置(下文)

一言不合就上个Demo:


+++++++++++++++++++++++++++++++++++++++++++++++++++++++++


1、建表:


create table test_trancaction
(user_id Int,name String)
clustered by (user_id) into 3 buckets
stored as orc TBLPROPERTIES ('transactional'='true');

2、导入数据:


create table test_insert_test(id int,name string) row format delimited fields TERMINATED BY ','; ---临时表
hive> select * from test_insert_test;
OK
1jerrick
2tom
3jerry
4lily
5hanmei
6limlei
7lucky
insert into test_trancaction select * from test_insert_test;
hive>
> select * from test_trancaction;
OK
1jerrick
2tom
3jerry
4lily
5hanmei
6limlei
7lucky

3、查看分桶情况:


hive >dfs-ls/user/hive/warehouse/test_trancaction/;
Found1items
-rwxrwxr-x3hadoophadoop 3522016-07-1316:24/user/hive/warehouse/test_trancaction/000000_0

并没有分桶,设置参数:set hive.enforce.bucketing=true


再次导入,查看日志:Numberofreducetasksdeterminedatcompiletime:3,再次check hdf文件,发现已经分成了3块。

4、尝试update:


修改配置hive-site




hive.support.concurrency
true


hive.enforce.bucketing
true


hive.exec.dynamic.partition.mode
nonstrict


hive.txn.manager
org.apache.hadoop.hive.ql.lockmgr.DbTxnManager


hive.compactor.initiator.on
true


hive.compactor.worker.threads
1

见证奇迹的时刻: updatetest_trancactionsetname='jerrick_up'whereid=1;


哪有那么一帆风顺,看看遇到的问题:


1、Load Data:


loaddatalocalinpath'/home/hadoop/wangjiankui/hives/transaction.csv'intotabletest_trancaction;


查询失败:


select*fromtest_trancaction;


Failedwithexceptionjava.io.IOException:java.io.IOException:MalformedORCfilehdfs://yq01-ns1/user/hive/warehouse/test_trancaction/transaction.csv.Invalidpostscript.


原因:load data是纯拷贝,不会做其他的处理,所以会报格式问题。


http://stackoverflow.com/questions/21721251/loading-data-from-a-txt-file-to-table-stored-as-orc-in-hive


2、连接metastore失败


select*fromtest_trancaction;


FAILED:LockException[Error10280]:Errorcommunicatingwiththemetastore

看日志:


Causedby:org.apache.thrift.TApplicationException:Internalerrorprocessingget_open_txns


重启metastore


Errorrollingback:Can'tcallrollbackwhenautocommit=true


2016-07-1318:11:13,056ERRORmetastore.HiveMetaStore(HiveMetaStore.java:run(6104))-Failurewhenstartingthecompactor,compactionsmaynothappen,java.lang.RuntimeException:Unabletosetuptransactiondatabasefortesting:Can'tcallrollbackwhenautocommit=true


不要加这个配置!!!!



hive.in.test


true


++++ 分割线(下面是官方wiki,重点标红)+++++++++++
Hive Transactions
ACID and Transactions in Hive
What is ACID and why should you use it?

ACID stands for four traits of database transactions: Atomicity (an operation either succeeds completely or fails, it does not leave partial data), Consistency (once an application performs an operation the results of that operation are visible to it in every subsequent operation), Isolation (operations by one user do not cause unexpected side effects for other users), and Durability (once an operation is complete it will be preserved even in the face of machine or system failure). These traits have long been expected of database systems as part of their transaction functionality.


Up until Hive 0.13, atomicity, consistency, and durability were provided at the partition level. Isolation could be provided by turning on one of the available locking mechanisms (ZooKeeperor in memory). With the addition of transactions inHive 0.13it is now possible to provide full ACID semantics at the row level, so that one application can add rows while another reads from the same partition without interfering with each other.


ACID是数据库事务性的4个基本要素:原子性、一致性、隔离性和持久性。


Transactions with ACID semantics have been added to Hive to address the following use cases:

Streaming ingest of data. Many users have tools such asApache Flume,Apache Storm, orApache Kafkathat they use to stream data into their Hadoop cluster. While these tools can write data at rates of hundreds or more rows per second, Hive can only add partitions every fifteen minutes to an hour. Adding partitions more often leads quickly to an overwhelming number of partitions in the table. These tools could stream data into existing partitions, but this would cause readers to get dirty reads (that is, they would see data written after they had started their queries) and leave many small files in their directories that would put pressure on the NameNode. With this new functionality this use case will be supported while allowing readers to get a consistent view of the data and avoiding too many files.
Slow changing dimensions. In a typical star schema data warehouse, dimensionstables change slowly over time. For example, a retailer will open new stores, which need to be added to the stores table, or an existing store may change its square footage or some other tracked characteristic. These changes lead to inserts of individual records or updatesof records (depending on the strategy chosen). Starting with 0.14, Hive is able to support this.
Data restatement. Sometimes collected data is found to be incorrect and needs correction. Or the first instance of the data may be an approximation (90% of servers reporting) with the full data provided later. Or business rules may require that certain transactions be restated due to subsequent transactions (e.g., after making a purchase a customer may purchase a membership and thus be entitled to discount prices, including on the previous purchase). Or a user may be contractually required to remove their customer’s data upon termination of their relationship. Starting with Hive 0.14 these use cases can be supported viaINSERT, UPDATE, andDELETE. LimitationsBEGIN,COMMIT, andROLLBACKare not yet supported. All language operations are auto-commit. The plan is to support these in a future release.
OnlyORC file formatis supported in this first release. The feature has been built such that transactions can be used by any storage format that can determine how updates or deletes apply to base records (basically, that has an explicit or implicit row id), but so far the integration work has only been done for ORC.

只支持ORCFile。

By default transactions are configured to be off. See theConfigurationsection below for a discussion of which values need to be set to configure it.

默认关闭,需要配置开启。

Tables must bebucketedto make use of these features. Tables in the same system not using transactions and ACID do not need to be bucketed.

表必须分桶。In general, distributing rows based on thehashwill give youa even distribution(均匀分布)in the buckets.


分桶介绍:http://www.aahyhaa.com/archives/316

At this time only snapshot level isolation is supported. When a given query starts it will be provided with a consistent snapshot of the data. There is no support for dirty read, read committed, repeatable read, or serializable. With the introduction of BEGIN the intention is to supportsnapshot isolation for the duration of transaction rather than just a single query. Other isolation levels may be added depending on user requests.

只支持快照级别的隔离。

The existing ZooKeeper and in-memory lock managers are not compatible with transactions. There is no intention to address this issue. SeeBasic Designbelow for a discussion of how locks are stored for transactions.
Schema changes using ALTER TABLE is NOT supported for ACID tables. HIVE-11421is tracking it.Fixed in 1.3.0/2.0.0.
Using Oracle as the Metastore DB and "datanucleus.connectionPoolingType=BONECP" may generate intermittent "No such lock.." and "No such transaction..." errors. Setting "datanucleus.connectionPoolingType=DBCP" is recommended in this case. Streaming APIs

Hive offers APIs for streaming data ingest and streaming mutation:

Hive HCatalog Streaming API
HCatalog Streaming Mutation API(available in Hive 2.0.0 and later)

A comparison of these two APIs is available in theBackgroundsection of the Streaming Mutation document.


Grammar Changes

INSERT...VALUES, UPDATE, andDELETEhave been added to the SQL grammar, starting inHive 0.14. SeeLanguageManual DMLfor details.


从0.14开始,支持INSET、UPDATE和DELETE.


Several new commands have been added to Hive's DDL in support of ACID and transactions, plus some existing DDL has been modified.


A new commandSHOW TRANSACTIONShas been added, seeShow Transactionsfor details.


A new commandSHOW COMPACTIONShas been added, seeShow Compactionsfor details.


TheSHOW LOCKScommand has been altered to provide information about the new locks associated with transactions. If you are using the ZooKeeper or in-memory lock managers you will notice no difference in the output of this command. SeeShow Locksfor details.


A new option has been added toALTER TABLEto request a compaction of a table or partition. In general users do not need to request compactions, as the system will detect the need for them and initiate the compaction. However, ifcompaction is turned offfor a table or a user wants to compact the table at a time the system would not choose to,ALTER TABLEcan be used to initiate the compaction. SeeAlter Table/Partition Compactfor details. This will enqueue a request for compaction and return. To watch the progress of the compaction the user can useSHOW COMPACTIONS.


Basic Design

HDFS does not support in-place changes to files. It also does not offer read consistency in the face of writers appending to files being read by a user. In order to provide these features on top of HDFS we have followed the standard approach used in other data warehousing tools. Data for the table or partition is stored in a set of base files. New records, updates, and deletes are stored in delta files. A new set of delta files is created for each transaction (or in the case of streaming agents such as Flume or Storm, each batch of transactions) that alters a table or partition. At read time the reader merges the base and delta files, applying any updates and deletes as it reads.


Delta File Compaction

Occasionally these changes need to be merged into the base files. To do this a set of threads has been added to the Hive metastore. They determine when this compaction needs to be done, execute the compaction, and then clean up afterwards. There are two types of compactions, minor and major.

Minor compaction takes a set of existing delta files and rewrites them to a single delta file per bucket.
Major compaction takes one or more delta files and the base file for the bucket and rewrites them into a new base file per bucket.

All compactions are done in the background and do not prevent concurrent reads and writes of the data. After a compaction the system waits until all readers of the old files have finished and then removes the old files.


Compactions are MR jobs with name in the following form: -compactor-.

.


Base and Delta Directories

Previously all files for a partition (or a table if the table is not partitioned) lived in a single directory. With these changes, any partitions (or tables) written with an ACID aware writer will have a directory for the base files and a directory for each set of delta files.


Lock Manager

A new lock manager has also been added to Hive, the DbLockManager. This lock manager stores all lock information in the metastore. In addition all transactions are stored in the metastore. This means that transactions and locks are durable in the face of server failure. To avoid clients dying and leaving transaction or locks dangling, a heartbeat is sent from lock holders and transaction initiators to the metastore on a regular basis. If a heartbeat is not received in the configured amount of time, the lock or transaction will be aborted.


As ofHive 1.3.0, the length of time that the DbLockManger will continue to try to acquire locks can be controlled viahive.lock.numretiresandhive.lock.sleep.between.retries. When the DbLockManager cannot acquire a lock (due to existence of a competing lock), it will back off and try again after a certain time period. In order to support short running queries and not overwhelm the metastore at the same time, the DbLockManager will double the wait time after each retry. The initial back off time is 100ms and is capped by hive.lock.sleep.between.retries. hive.lock.numretries is the total number of times it will retry a given lock request. Thus the total time that the call to acquire locks will block (given default values of 10 retries and 60s sleep time) is (100ms + 200ms + 400ms + ... + 51200ms + 60s + 60s + ... + 60s) = 91m:42s:300ms.


Moredetailson locks used by this Lock Manager.


Configuration

These configuration parameters must be set appropriately to turn on transaction support in Hive:

hive.support.concurrency– true
hive.enforce.bucketing– true (Not required as ofHive 2.0)
hive.exec.dynamic.partition.mode– nonstrict
hive.txn.manager–org.apache.hadoop.hive.ql.lockmgr.DbTxnManager
hive.compactor.initiator.on–true (for exactly one instance of the Thrift metastore service)
hive.compactor.worker.threads– a positive numberon at least one instance of the Thrift metastore service

The following sections list all of the configuration parameters that affect Hive transactions and compaction. Also seeLimitationsabove andTable Propertiesbelow.


New Configuration Parameters for Transactions

A number of new configuration parameters have been added to the system to support transactions.



Configuration key


Values


Location

Notes




hive.txn.manager


Default:org.apache.hadoop.hive.ql.lockmgr.DummyTxnManager

Value required for transactions:org.apache.hadoop.hive.ql.lockmgr.DbTxnManager


Client/ HiveServer2


DummyTxnManager replicates pre Hive-0.13 behavior and provides no transactions.




hive.txn.timeout


Default:300


Client/ HiveServer2/

Metastore


Time after which transactions are declared aborted if the client has not sent a heartbeat, in seconds. It's critical that this property has the same value for all components/services.5




hive.timedout.txn.reaper.start
Default:100s
Metastore
Time delay of first reaper (the process which aborts timed-out transactions) run after the metastore starts (as ofHive 1.3.0).


hive.timedout.txn.reaper.interval


Default:180s
Metastore

Time interval describing how often the reaper (the process which aborts timed-out transactions) runs (as ofHive 1.3.0).




hive.txn.max.open.batch


Default:1000


Client

Maximum number of transactions that can be fetched in one call to open_txns().1




hive.compactor.initiator.on


Default:false

Value required for transactions:true (for exactly one instance of the Thrift metastore service)


Metastore

Whether to run the initiator and cleaner threads on this metastore instance.It's critical that this is enabled on exactly one metastore service instance (not enforced yet).

As ofHive 1.3.0this property may be enabled on any number of metastore instances.




hive.compactor.worker.threads


Default:0

Value required for transactions:> 0 on at least one instance of the Thrift metastore service


Metastore

How many compactor worker threads to run on this metastore instance.2




hive.compactor.worker.timeout


Default:86400


Metastore

Time in seconds after which a compaction job will be declared failed and the compaction re-queued.




hive.compactor.cleaner.run.interval
Default: 5000
Metastore
Time in milliseconds between runs of the cleaner thread. (Hive 0.14.0and later.)


hive.compactor.check.interval


Default:300


Metastore

Time in seconds between checks to see if any tables or partitions need to be compacted.3




hive.compactor.delta.num.threshold


Default:10


Metastore

Number of delta directories in a table or partition that will trigger a minor compaction.




hive.compactor.delta.pct.threshold


Default:0.1


Metastore

Percentage (fractional) size of the delta files relative to the base that will trigger a major compaction. 1 = 100%, so the default 0.1 = 10%.




hive.compactor.abortedtxn.threshold


Default:1000


Metastore

Number of aborted transactions involving a given table or partition that will trigger a major compaction.




hive.compactor.max.num.delta


Default: 500
Metastore
Maximum number of delta files that the compactor will attempt to handle in a single job(as ofHive 1.3.0).4


hive.compactor.job.queue


Default: "" (empty string)
Metastore
Used to specify name of Hadoop queue to which Compaction jobs will be submitted. Set to empty string to let Hadoop choose the queue(as ofHive 1.3.0).

1hive.txn.max.open.batchcontrols how many transactions streaming agents such as Flume or Storm open simultaneously. The streaming agent then writes that number of entries into a single file (per Flume agent or Storm bolt). Thus increasing this value decreases the number of delta files created by streaming agents. But it also increases the number of open transactions that Hive has to track at any given time, which may negatively affect read performance.


2Worker threads spawn MapReduce jobs to do compactions. They do not do the compactions themselves. Increasing the number of worker threads will decrease the time it takes tables or partitions to be compacted once they are determined to need compaction. It will also increase the background load on the Hadoop cluster as more MapReduce jobs will be running in the background.


3Decreasing this value will reduce the time it takes for compaction to be started for a table or partition that requires compaction. However, checking if compaction is needed requires several calls to the NameNode for each table or partition that has had a transaction done on it since the last major compaction. So decreasing this value will increase the load on the NameNode.


4If the compactor detects a very high number of delta files, it will first run several partial minor compactions (currently sequentially) and then perform the compaction actually requested.


5If the value is not the same active transactions may be determined to be "timed out" and consequently Aborted. This will result in errors like "No such transaction...", "No such lock ..."


Configuration Values to Set forINSERT, UPDATE, DELETE

In addition to the new parameters listed above, some existing parameters need to be set to supportINSERT ... VALUES, UPDATE,andDELETE.





hive.support.concurrency
true (default is false)


hive.enforce.bucketing
true (default is false) (Not required as ofHive 2.0)


hive.exec.dynamic.partition.mode
nonstrict (default is strict)

Configuration Values to Set for Compaction

If the data in your system is not owned by the Hive user (i.e., the user that the Hive metastore runs as), then Hive will need permission to run as the user who owns the data in order to perform compactions. If you have already set up HiveServer2 to impersonate users, then the only additional work to do is assure that Hive has the right to impersonate users from the host running the Hive metastore. This is done by adding the hostname tohadoop.proxyuser.hive.hostsin Hadoop'score-site.xmlfile. If you have not already done this, then you will need to configure Hive to act as a proxy user. This requires you to set up keytabs for the user running the Hive metastore and addhadoop.proxyuser.hive.hostsandhadoop.proxyuser.hive.groupsto Hadoop'score-site.xmlfile. See the Hadoop documentation on secure mode for your version of Hadoop (e.g., for Hadoop 2.5.1 it is atHadoop in Secure Mode).


Table Properties

If a table is to be used in ACID writes (insert, update, delete) then the table property "transactional=true"must be set on that table, starting withHive 0.14.0. Also,hive.txn.managermust be set to org.apache.hadoop.hive.ql.lockmgr.DbTxnManager either in hive-site.xml or in the beginning of the session before any query is run.Without those, inserts will be done in the old style; updates and deletes will be prohibited. However, this does not apply to Hive 0.13.0.


If a table owner does not wish the system to automatically determine when to compact, then the table property "NO_AUTO_COMPACTION" can be set. This will prevent all automatic compactions. Manual compactions can still be done withAlter Table/Partition Compactstatements.


Table properties are set with the TBLPROPERTIES clause when a table is created or altered, as described in theCreate TableandAlter Table Propertiessections of Hive Data Definition Language. The "transactional" and "NO_AUTO_COMPACTION" table properties are case-sensitive in Hive releases 0.x and 1.0, but they are case-insensitivestarting with release 1.1.0 (HIVE-8308).


最新文章

123

最新摄影

微信扫一扫

第七城市微信公众平台

Configuration key

Must be set to