-
Notifications
You must be signed in to change notification settings - Fork 1.6k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
GH-2852: support fixed transaction id suffix #2913
Conversation
add properties `maxCache` at `DefaultKafkaProducerFactory`, setting `maxCache` greater than zero can reuse `transactional.id`. Resolves spring-projects#2852
} | ||
} | ||
else { | ||
suffix = String.valueOf(this.transactionIdSuffix.getAndIncrement()); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Nice work, and thanks for your effort! However, I think just incrementing may not be sufficient. The goal of GH-2852 is to allow users to set how transactionIds
are issued. For example, with a topic having 9 partitions, 3 replicas of my service (each with 3 producers), and limits on 9 txIds
, restarting my service becomes a bit of a pain. It takes time to distribute transaction IDs correctly(without getting Fenced
all the time), so instead of that, I would prefer to use external storage as a source of truth.
I suggest introducing a new interface for generating transactionIdSuffix
with a default increment implementation. WDYT?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I suggest introducing a new interface for generating transactionIdSuffix with a default increment implementation. WDYT?
Via new interface, the first instance of service increments transactionIdSuffix
from 0 and the second instance of service increments from 3?
If I understand correctly, I like it.
At the same time we need to monitor the generated trasaction.id
, such as provide method return current transactionIdSuffix
or pushlish event carry curr transactionIdSuffix
at DefaultKafkaProducerFactory.destroy()
.
As an alternative, write a unique ID in the transactionIdPrefix
, such as UUD is feasible?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Via the new interface, does the first instance of the service increment transactionIdSuffix from 0, and the second instance increments it from 3?
It actually depends on the implementation. It could distribute transactionIdSuffix
between replicas as described, or it could provide the first free suffix from the source of truth storage
. However, the default implementation remains the same as it is now.
At the same time, we need to monitor the generated transaction.id. For example, provide a method to return the current transactionIdSuffix or publish an event carrying the current transactionIdSuffix at DefaultKafkaProducerFactory.destroy().
Certainly, we need this functionality. Therefore, the interface must include a contract to retrieve the transactionId
and return it back.
As an alternative, writing a unique ID in the transactionIdPrefix, such as UUID, is feasible?
In my environment, it's not an option because I have a mask for transactionIds
like topic-name-{0..n}(e.g. mytopic-1).
This comment was marked as abuse.
This comment was marked as abuse.
Sorry, something went wrong.
Introducing a new interface for generating transactionIdSuffix with a default increment implementation. Introducing a new interface for return the current transactionIdSuffix. Resolves spring-projects#2852
Hi @Wzy19930507, I am taking a look at this PR. Could you let me know if you added the new interface that @stillya suggested? I don't see that in the PR yet. |
|
||
Since 3.1, when setting `maxCache` greater than zero can reuse `transactional.id` within a specific range. | ||
When a transaction producer is requested and `transactional.id` all in use, throw a `NoProducerAvailableException`. | ||
User can use then use a RetryTemplate configured to retry that exception, with a suitably configured back off. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Change the wording to - "We can then use a"
When setting `maxCache` to 5, `transactional.id` is `my.txid.`++`{0-4}`+. | ||
|
||
IMPORTANT: When use `KafkaTransactionManager` in the `ConcurrentMessageListenerContainer`, `maxCache` must be greater than `concurrency`, also be careful nested transaction. | ||
|
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
"also be careful about".
It would be beneficial if you could add a few sentences detailing why the maxCache
needs to be greater than concurrency
. Same thing about nested transactions.
public final boolean initTransactionIdSuffix(int transactionIdSuffix) { | ||
Assert.isTrue(transactionIdSuffix >= 0, "'transactionIdSuffix' initial value must greater than or equal 0"); | ||
return this.transactionIdSuffix.compareAndSet(0, transactionIdSuffix); | ||
} |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Hi, @sobychacko new interface is here. DefaultKafkaProducerFactory
use only
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thank you, but I didn't quite mean that, so I've decided to lend a hand. Could you please take a look?
Add properties
maxCache
atDefaultKafkaProducerFactory
, settingmaxCache
greater than zero can reusetransactional.id
.If maxCache is 3, the suffix corresponding to the first prefix is 0-2, corresponding to the second prefix is 3-5.
Add new interface to initialize transactionIdSuffix.
Resolves #2852