I have a bank project which customer balances should be updated by parallel threads in parallel applications. I hold customer balances in an Oracle database. My java applications will be implemented with Spring and Hibernate.
How can i implement the race condition between parallel applications? Should my solution be at database level or at application level?
I assume what you would like to know is how to handle concurrency, preventing race conditions which can occur where two parts of the application modify and accidentally overwrite the same data.
You have mostly two strategies for this: pessimistic locking and optimistic locking:
Pessimistic locking
here you assume that the likelyhood that two threads overwrite the same data is high, so you would like it to handle it in a transparent way. To handle this, increase the isolation level of your Spring transactions from it's default value of READ_COMMITTED to for example REPEATABLE_READ which should be sufficient in most cases:
#Transactional(isolation=Isolation.REPEATABLE_READ)
public void yourBusinessMethod {
...
}
In this case if you read some data in the beginning of the method, you are sure that noone can overwrite the data in the database while your method is ongoing. Note that it's still possible for another thread to insert extra records to a query you made (a problem known as phantom reads), but not change the records you already read.
If you want to protect against phantom reads, you need to upgrade the isolation level to SERIALIZABLE. The improved isolation comes at a performance cost, your program will run slower and will more frequently 'hang' waiting for the other part of the program to finish.
Optimistic Locking
Here you assume that data access colisions are rare, and that in the rare cases they occur they are easilly recoverable by the application. In this mode, you keep all your business methods in their default REPEATABLE_READ mode.
Then each Hibernate entity is marked with a version column:
#Entity
public SomeEntity {
...
#Version
private Long version;
}
With this each entity read from the database is versioned using the version column. When Hibernate write changes to an entity in the database, it will check if the version was incremented since the last time that transaction read the entity.
If so it means someone else modified the data, and decisions where made using stale data. In this case a StaleObjectException is thrown, that needs to be caught by the application and handled, ideally at a central place.
In the case of a GUI, you usuall catch the exception, show a message saying user xyz changed this data while you where also editing it, your changes are lost. Press Ok to reload the new data.
With optimistic locking your program will run faster but the applications needs to handle some concurrency aspects that would otherwise be transparent with pessimistic locking: version entities, catch exceptions.
The most frequently used method is optimistic locking, as it seems to be acceptable in most applications. With pessimistic locking it's very easy to cause performance problems, specially when data access colisions are rare and can be solved in a simple way.
There are no constraints to mix the use of the two concurrency handling methods in the same application if needed.
Related
I am a little confused as to why Optimistic Locking is actually safe. If I am checking the version at the time of retrieval with the version at the time of update, it seems like I can still have two requests enter the update block if the OS issues an interrupt and swaps the processes before the commit actually occurs. For example:
latestVersion = vehicle.getVersion();
if (vehicle.getVersion() == latestVersion) {
// update record in database
} else {
// don't update record
}
In this example, I am trying to manually use Optimistic Locking in a Java application without using JPA / Hibernate. However, it seems like two requests can enter the if block at the same time. Can you please help me understand how to do this properly? For context, I am also using Java Design Patterns website as an example.
Well... that's the optimistic part. The optimism is that it is safe. If you have to be certain it's safe, then that's not optimistic.
The example you show definitely is susceptible to a race condition. Not only because of thread scheduling, but also due to transaction isolation level.
A simple read in MySQL, in the default transaction isolation level of REPEATABLE READ, will read the data that was committed at the time your transaction started.
Whereas updating data will act on the data that is committed at the time of the update. If some other concurrent session has updated the row in the database in the meantime, and committed it, then your update will "see" the latest committed row, not the row viewed by your get method.
The way to avoid the race condition is to not be optimistic. Instead, force exclusive access to the record. Doveryai, no proveryai.
If you only have one app instance, you might use a critical section for this.
If you have multiple app instances, critical sections cannot coordinate other instances, so you need to coordinate in the database. You can do this by using pessimistic locking. Either read the record using a locking read query, or else you can use MySQL's user-defined locks.
I have a problem concerning java optimistic locking exception. I have a service class that is instantiated (by spring) for every new user session and it contains a non static method that perform db operations. I wonder how I can avoid optimistic locking exception on the entity that is read/written to db. I would like to achieve a similar result as a synchronized method would but I guess using "synchronized" is out of the question since the method is not static and would not have any effect when users have own instances of the service? Can I somehow detect if a new version of the entity is saved to db and then retrieve a new version and then edit and save that one? I want the transaction to hold until it is ok even if it implies the transaction have to wait for other transactions. My first idea was to put the transaction code into a try-catch block and then retry the transaction (read & write) if optimistic locking exceptions is thrown. Is that solution "too easy" or?
Optimistic locking is used to improve performance, but still avoid messing up the data.
If there's an Optimistic lock failure, the user (that failed the update) needs to decide if he wants to do his operation again. You can't automate that, since it depends entirely on what was changed and how.
So no, your idea of a retry the transaction with a try/catch is not a "too easy solution". It's not a solution, it would be a serious (and dumb) bug.
I have a method in which some database insert operations are happening using hibernate and i want them to be thread safe. The method is getting some data in parametres and its a possiblity that sometimes two calls are made with same data at same point of time.
I can't lock those tables because of performance degradation. Can anyone suggest making the method as synchronized will solve issue?
Synchronizing a method will ensure that it can only be accessed by one thread at a time. If this method is your only means of writing to the database, then yes, this will stop two threads from writing at the same time. However, you still have to deal with the fact that you have multiple insert operations with the same data.
You should let Hibernate handle the concurrency, that's what it is meant to do. Don't assume Hibernate will lock anything: it supports optimistic transactions for exactly this purpose. Quote from the above link:
The only approach that is consistent with high concurrency and high scalability, is optimistic concurrency control with versioning. Version checking uses version numbers, or timestamps, to detect conflicting updates and to prevent lost updates. Hibernate provides three possible approaches to writing application code that uses optimistic concurrency.
Database Concurrency is handled by transactions. Transactions have the Atomic Consistent Isolated Durable (ACID) properties. They provide isolation between programs accessing a database concurrently. In the Hibernate DAO template of spring framework there are single line methods for CRUD operations on the database. When used individually these don't need to be synchronized by method. Spring provides declarative (XML), programmatic and annotation meta-data driven transaction management if you need to declare "your method" as transactional with specific propagation settings, rollbackFor settings, isolation settings. So in "your method" you can do multiple save,update,deletes etc and the ORM will ensure that it is executed with the transaction settings you have given in the meta-data.
Another issue is that the thread has to have the lock on all the objects that are taking part in the transaction.Otherwise the transaction might fail or the ORM will persist stale data. In another situation it can result in a deadlock because of lock-ordering. I think this is what really answers your question.
Both objects a and b have an instance variable of the type Lock. A boolean flag can be used to indicate the success of the transaction. The client code can retry the same transaction if it fails.
if (a.lock.tryLock()) {
try {
if (b.lock.tryLock()) {
try {
// persist or update object a and b
} finally {
b.lock.unlock();
}
}
} finally {
a.lock.unlock();
}
}
The problem with using synchronized methods is that it locks up the entire Service or DAO class making other service methods unavailable to other threads. By using individual locks on objects we can gain the advantage of fine grained concurrency.
No. This method probably uses another methods and objects, which may be not thread safe. synchronized makes threads to use that's method's object monitor only once at a time, so it makes thread-safe a method with respect to the object.
If you are sure that all other threads use shared functionality only with this method, then making it synchronized may be sufficient.
Choosing the best strategy depends on the architecture, sometimes to increase performance seems to be easier to use the trick like method synchronization, but this is bad approach.
There's no doubts, you should use transactions, and if with that strategy you're facing performance issues you should optimize your db queries or db structure.
Please remember that "Synchronization" should be as much as possible atomic.
I am looking at EntityManager API, and I am trying to understand an order in which I would do a record lock. Basically when a user decides to Edit a record, my code is:
entityManager.getTransaction().begin();
r = entityManager.find(Route.class, r.getPrimaryKey());
r.setRoute(txtRoute.getText());
entityManager.persist(r);
entityManager.getTransaction().commit();
From my trial and error, it appears I need to set WWEntityManager.entityManager.lock(r, LockModeType.PESSIMISTIC_READ); after the .begin().
I naturally assumed that I would use WWEntityManager.entityManager.lock(r, LockModeType.NONE); after the commit, but it gave me this:
Exception Description: No transaction is currently active
I haven't tried putting it before the commit yet, but wouldn't that defeat the purpose of locking the record, since my goal is to avoid colliding records in case 50 users try to commit a change at once?
Any help as to how to I can lock the record for the duration of the edit, is greatly appreciated!
Thank You!
Performing locking inside transaction makes perfectly sense. Lock is automatically released in the end of the transaction (commit / rollback). Locking outside of transaction (in context of JPA) does not make sense, because releasing lock is tied to end of the transaction. Also otherwise locking after changes are performed and transaction is committed does not make too much sense.
It can be that you are using pessimistic locking to purpose other than what they are really for. If my assumption is wrong, then you can ignore end of the answer. When your transaction holds pessimistic read lock on entity (row), following is guaranteed:
No dirty reads: other transactions cannot see results of operations you performed to locked rows.
Repeatable reads: no modifications from other transactions
If your transaction modifies locked entity, PESSIMISTIC_READ is upgraded to PESSIMISTIC_WRITE or transaction fails if lock cannot be upgraded.
Following coarsely describes scenario with obtaining locking in the beginning of transaction:
entityManager.getTransaction().begin();
r = entityManager.find(Route.class, r.getPrimaryKey(),
LockModeType.PESSIMISTIC_READ);
//from this moment on we can safely read r again expect no changes
r.setRoute(txtRoute.getText());
entityManager.persist(r);
//When changes are flushed to database, provider must convert lock to
//PESSIMISTIC_WRITE, which can fail if concurrent update
entityManager.getTransaction().commit();
Often databases do not have separate support for pessimistic read, so you are actually holding lock to row since PESSIMISTIC_READ. Also using PESSIMISTIC_READ makes sense only if no changes to the locked row are expected. In case above changes are done always, so using PESSIMISTIC_WRITE from the beginning on is reasonable, because it saves you from the risk of concurrent update.
In many cases it also makes sense to use optimistic instead of pessimistic locking. Good examples and some comments about choosing between locking strategies can be found from: Locking and Concurrency in Java Persistence 2.0
Great work attempting to be safe in write locking your changing data. :) But you might be going overboard / doing it the long way.
First a minor point. The call to persist() isn't needed. For update, just modify the attributes of the entity returned from find(). The entityManager automatically knows about the changes and writes them to the db during commit. Persist is only needed when you create a new object & write it to the db for the first time (or add a new child object to a parent relation and which to cascade the persist via cascade=PERSIST).
Most applications have a low probability of 'clashing' concurrent updates to the same data by different threads which have their own separate transactions and separate persistent contexts. If this is true for you and you would like to maximise scalability, then use an optimistic write lock, rather than a pessimistic read or write lock. This is the case for the vast majority of web applications. It gives exactly the same data integrity, much better performance/scalability, but you must (infrequently) handle an OptimisticLockException.
optimistic write locking is built-in automatically by simply having a short/integer/long/TimeStamp attribute in the db and entity and annotating it in the entity with #Version, you do not need to call entityManager.lock() in that case
If you were satisfied with the above, and you added a #Version attribute to your entity, your code would be:
try {
entityManager.getTransaction().begin();
r = entityManager.find(Route.class, r.getPrimaryKey());
r.setRoute(txtRoute.getText());
entityManager.getTransaction().commit();
} catch (OptimisticLockException e) {
// Logging and (maybe) some error handling here.
// In your case you are lucky - you could simply rerun the whole method.
// Although often automatic recovery is difficult and possibly dangerous/undesirable
// in which case we need to report the error back to the user for manual recovery
}
i.e. no explicit locking at all - the entity manager handles it automagically.
IF you had a strong need to avoid concurrent data update "clashes", and are happy to have your code with limited scalability then serialise data access via pessimistic write locking:
try {
entityManager.getTransaction().begin();
r = entityManager.find(Route.class, r.getPrimaryKey(), LockModeType.PESSIMISTIC_WRITE);
r.setRoute(txtRoute.getText());
entityManager.getTransaction().commit();
} catch (PessimisticLockException e) {
// log & rethrow
}
In both cases, a successful commit or an exception with automatic rollback means that any locking carried out is automatically cleared.
Cheers.
I was going through ACID properties regarding Transaction and encountered the statement below across the different sites
ACID is the acronym for the four properties guaranteed by transactions: atomicity, consistency, isolation, and durability.
**My question is specifically about the phrase.
guaranteed by transactions
**. As per my experience these properties are not taken care by
transaction automatically. But as a java developer we need to ensure that these properties criteria are met.
Let's go through for each property:-
Atomicity:- Assume when we create the customer the account should be created too as it is compulsory. So now during transaction
the customer gets created while during account creation some exception oocurs. So the developer can now go two ways: either he rolls back the
complete transaction (atomicity is met in this case) or he commits the transaction so customer will be created but not the
account (which violates the atomicity). So responsibility lies with developer?
Consistency:- Same reason holds valid for consistency too
Isolation :- as per definition isolation makes a transaction execute without interference from another process or transactions.
But this is achieved when we set the isolation level as Serializable. Otherwis in another case like read commited or read uncommited
changes are visible to other transactions. So responsibility lies with the developer to make it really isolated with Serializable?
Durability:- If we commit the transaction, then even if the application crashes, it should be committed on restart of application. Not sure if it needs to be taken care by developer or by database vendor/transaction?
So as per my understanding these ACID properties are not guaranteed automatically; rather we as a developer sjould achieve them. Please let me know
if above understanding regarding each point is correct? Would appreciate if you folks can reply for each point(yes/no will also do.
As per my understanding read committed should be most logical isolation level in most application, though it depends on requirement too.
The transactions guarantees ACID more or less:
1) Atomicity. Transaction guarantees all changes are made or none of them. But you need to manually set the start and end of a transaction and manually perform commit or rollback. Depending on the technology you use (EJB...), transactions are container-managed, setting the start and end to the whole "method" you are creating. You can control by configuration if a method invoked requires a new transaction or an existing one, no transaction...
2) Consistency. Guaranteed by atomicity.
3) Isolation. You must define the isolation level your application needs. Default value is defined depending upon the database, container... The commonest one is READ COMMITTED. Be careful with locks as can cause dead-lock depending on your logic and isolation level.
4) Durability. Managed entirely by the database. If your commit executes without error, nearly all database guarantees durability of changes, but some scenarios can cause to not guarantee that (writes to disk are cached in memory and flushed later...)
In general, you should be aware of transactions and configure it in the container of declare by code the star and end (commit, rollback).
Database transactions are atomic: They either happen in their entirety or not at all. By itself, this says nothing about the atomicity of business transactions. There are various strategies to map business transactions to database transactions. In the simplest case, a business transaction is implemented by one database transaction (where a business transaction is aborted by rolling back the database one). Then, atomicity of database transactions implies atomicity of business transactions. However, things get tricky once business transactions span several database transactions ...
See above.
Your statement is correct. Often, the weaker guarantees are sufficient to prove correctness.
Database transactions are durable (unless there is a hardware failure): if the transaction has committed, its effect will persist until other transactions change the data. However, calling code might not learn whether a transaction has comitted if the database or the network between database and calling code fails. Therefore
If we commit the transaction, then even if application crash, it should be committed on restart of application.
is wrong. If the transaction has committed, there is nothing left to do.
To summarize, the database does give strong guarantees - about the behaviour of the database. Obviously, it can not give guarantees about the behaviour of the entire application.