Hi, I've updated the KIP with the new #stop(boolean isDeleted) overloaded method, and have also amended the PR and JIRA tickets. I also added a couple entries to the "Rejected alternatives" section with the reasons why I pivoted from introducing new callback methods to retrofit the existing one.
Please let me know what your thoughts are. Cheers, Hector From: Hector Geraldino (BLOOMBERG/ 919 3RD A) At: 11/16/22 17:38:59 UTC-5:00To: dev@kafka.apache.org Subject: Re: [DISCUSS] KIP-883: Add delete callback method to Connector API Hi Mickael, I agree that the new STOPPED state proposed in KIP-875 will improve the connector lifecycle. The changes proposed in this KIP aim to cover the gap where connectors need to actually be deleted, but because the API doesn't provide any hooks, external assets are left lingering where they shouldn't. I agree that this proposal is similar to KIP-419, maybe the main difference is their focus on Tasks whereas KIP-833 proposes changes to the Connector. My goal is to figure out the correct semantics for notifying connectors that they're being stopped because the connector has been deleted. Now, computing the "deleted" state in both the Standalone and Distributed herders is not hard, so the question is: when shall the connector be notified? The "easiest" option would be to do it by calling an overloaded Connector#stop(deleted) method, but there are other - more expressive - ways, like providing an 'onDelete()' or 'destroy()' method. I'm leaning towards adding an overload method (less complexity, known corner cases), and will amend the KIP with the reasoning behind that decision soon. Thanks for your feedback! From: dev@kafka.apache.org At: 11/16/22 11:13:17 UTC-5:00To: dev@kafka.apache.org Subject: Re: [DISCUSS] KIP-883: Add delete callback method to Connector API Hi Hector, Thanks for the KIP. One tricky aspect is that currently there's no real way to stop a connector so to do so people often just delete them temporarily. KIP-875 proposes adding a mechanism to properly stop connectors which should reduce the need to deleting them and avoid doing potentially expensive cleanup operations repetitively. This KIP also reminds me of KIP-419: https://cwiki.apache.org/confluence/display/KAFKA/KIP-419%3A+Safely+notify+Kafka +Connect+SourceTask+is+stopped. Is it guaranteed the new delete callback will be the last method called? Thanks, Mickael On Tue, Nov 15, 2022 at 5:40 PM Sagar <sagarmeansoc...@gmail.com> wrote: > > Hey Hector, > > Thanks for the KIP. I have a minor suggestion in terms of naming. Since > this is a callback method, would it make sense to call it onDelete()? > > Also, the failure scenarios discussed by Greg would need handling. Among > other things, I like the idea of having a timeout for graceful shutdown or > else try a force shutdown. What do you think about that approach? > > Thanks! > Sagar. > > On Sat, Nov 12, 2022 at 1:53 AM Hector Geraldino (BLOOMBERG/ 919 3RD A) < > hgerald...@bloomberg.net> wrote: > > > Thanks Greg for taking your time to review not just the KIP but also the > > PR. > > > > 1. You made very valid points regarding the behavior of the destroy() > > callback for connectors that don't follow the happy path. After thinking > > about it, I decided to tweak the implementation a bit and have the > > destroy() method be called during the worker shutdown: this means it will > > share the same guarantees the connector#stop() method has. An alternative > > implementation can be to have an overloaded connector#stop(boolean deleted) > > method that signals a connector that it is being stopped due to deletion, > > but I think that having a separate destroy() method provides clearer > > semantics. > > > > I'll make sure to ammend the KIP with these details. > > > > 3. Without going too deep on the types of operations that can be performed > > by a connector when it's being deleted, I can imagine the > > org.apache.kafka.connect.source.SourceConnector base class having a default > > implementation that deletes the connector's offsets automatically > > (controlled by a property); this is in the context of KIP-875 (first-class > > offsets support in Kafka Connect). Similar behaviors can be introduced for > > the SinkConnector, however I'm not sure if this KIP is the right place to > > discuss all the possibilities, or if we shoold keeping it more > > narrow-focused on providing a callback mechanism for when connectors are > > deleted, and what the expectations are around this newly introduced method. > > What do you think? > > > > > > From: dev@kafka.apache.org At: 11/09/22 16:55:04 UTC-5:00To: > > dev@kafka.apache.org > > Subject: Re: [DISCUSS] KIP-883: Add delete callback method to Connector API > > > > Hi Hector, > > > > Thanks for the KIP! > > > > This is certainly missing functionality from the native Connect framework, > > and we should try to make it possible to inform connectors about this part > > of their lifecycle. > > However, as with most functionality that was left out of the initial > > implementation of the framework, the details are more challenging to work > > out. > > > > 1. What happens when the destroy call throws an error, how does the > > framework respond? > > > > This is unspecified in the KIP, and it appears that your proposed changes > > could cause the herder to fail. > > From the perspective of operators & connector developers, what is a > > reasonable expectation to have for failure of a destroy? > > I could see operators wanting both a graceful-delete to make use of this > > new feature, and a force-delete for when the graceful-delete fails. > > A connector developer could choose to swallow all errors encountered, or > > fail-fast to indicate to the operator that there is an issue with the > > graceful-delete flow. > > If the alternative is crashing the herder, connector developers may choose > > to hide serious errors, which is undesirable. > > > > 2. What happens when the destroy() call takes a long time to complete, or > > is interrupted? > > > > It appears that your implementation serially destroy()s each appropriate > > connector, and may prevent the herder thread from making progress while the > > operation is ongoing. > > We have previously had to patch Connect to perform all connector and task > > operations on a background thread, because some connector method > > implementations can stall indefinitely. > > Connect also has the notion of "cancelling" a connector/task if a graceful > > shutdown timeout operation takes too long. Perhaps some of that design or > > machinery may be useful to protect this method call as well. > > > > More specific to the destroy() call itself, what happens when a connector > > completes part of a destroy operation and then cannot complete the > > remainder, either due to timing out or a worker crashing? > > What is the contract with the connector developer about this method? Is the > > destroy() only started exactly once during the lifetime of the connector, > > or may it be retried? > > > > 3. What should be considered a reasonable custom implementation of the > > destroy() call? What resources should it clean up by default? > > > > I think we can broadly categorize the state a connector mutates among the > > following > > * Framework-managed state (e.g. source offsets, consumer offsets) > > * Implementation detail state (e.g. debezium db history topic, audit > > tables, temporary accounts) > > * Third party system data (e.g. the actual data being written by a sink > > connector) > > * Third party system metadata (e.g. tables in a database, delivery > > receipts, permissions) > > > > I think it's apparent that the framework-managed state cannot/should not be > > interacted with by the destroy() call. However, the framework could be > > changed to clean up these resources at the same time that destroy() is > > called. Is that out-of-scope of this proposal, and better handled by manual > > intervention? > > From the text of the KIP, I think it explicitly includes the Implementation > > detail state, which should not be depended on externally and should be safe > > to clean up during a destroy(). I think this is completely reasonable. > > Are the third-party data and metadata out-of-scope for this proposal? Can > > we officially recommend against it, or should we accommodate users and > > connector developers that wish to clean up data/metadata during destroy()? > > > > 4. How should connector implementations of destroy handle backwards > > compatibility? > > > > In terms of backward-compatibility for the framework vs connector versions, > > I think the default-noop method is very reasonable. > > However, what happens when someone upgrades from a version of a connector > > without a destroy() implementation to one with an implementation, and > > maintain backwards compatibility? > > To replicate the same behavior, the connector might include something like > > an `enable.cleanup` config which allows users to opt-in to the new > > behavior. This could mean the proliferation of many different > > configurations to handle this behavior. > > Maybe we can provide some recommendations to developers, or some mechanism > > to standardize this opt-in behavior. > > > > I'm interested to hear if you have any experience with the above, if you've > > experimented with this feature in your fork. > > > > Thanks, > > Greg > > > > > > On Thu, Nov 3, 2022 at 11:55 AM Hector Geraldino (BLOOMBERG/ 919 3RD A) < > > hgerald...@bloomberg.net> wrote: > > > > > Hi everyone, > > > > > > I've submitted KIP-883, which introduces a callback to the public > > > Connector API called when deleting a connector: > > > > > > > > > > > > > https://cwiki.apache.org/confluence/display/KAFKA/KIP-883%3A+Add+delete+callback > > +method+to+Connector+API > > > > > > It adds a new `deleted()` method (open to better naming suggestions) to > > > the org.apache.kafka.connect.connector.Connector abstract class, which > > will > > > be invoked by connect Workers when a connector is being deleted. > > > > > > Feedback and comments are welcome. > > > > > > Thank you! > > > Hector > > > > > > > > > > > >
