Don't consider deleted schema when checking compatibility #4669
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
feca5bb changed topic delete logic to delete the schema when the topic is deleted (though this only seems to be enabled for idle topic GC). This exposed a bug in compatibility checking whereby if the a subscription tries to attach to the topic, even if using the same schema as had been used previously, a compatibility exception will be thrown. The root cause was a logic bomb in the compatibility check. !(A || B) && C where it should have been !(A || B) || C. The issue doesn't affect producers because the check was already correct there. I've also updated the check for transitive compatibility to remove the prefix of schemas before the deleted schema. Previously this was throwing an NPE on the broker as it couldn't decode the deleted schema. This issue was discovered by failures in the healthcheck. The check period (5 minutes) was longer than the GC period (60 seconds). I would expect it to hit quite often in other scenarios also.
ivankelly
added
type/bug
|
CI won't pass until #4667 is merged. |
|
retest this please |
|
run cpp tests |
|
rerun java8 tests |
codelipenghui
approved these changes
Jul 4, 2019
|
rerun java8 tests |
jiazhai
approved these changes
Jul 4, 2019
|
rerun java8 tests |
|
run java8 tests |
The previous fix didn't take into account the case where if a producer is actively publishing to a topic, a consumer should not be allowed to attach with a schema.
|
@sijie @jiazhai @codelipenghui @merlimat you you take another look? I had to change the approach to strip the deleted schemas in getSchema, because a consumer should be prevented from attaching with a schema if there is a producer currently producing without a schema. |
|
rerun java8 tests |
|
rerun integration tests |
easyfan
pushed a commit
to easyfan/pulsar
that referenced
this pull request
Jul 26, 2019
feca5bb changed topic delete logic to delete the schema when the topic is deleted (though this only seems to be enabled for idle topic GC). This exposed a bug in compatibility checking whereby if the a subscription tries to attach to the topic, even if using the same schema as had been used previously, a compatibility exception will be thrown. This is because the topic still appears to have a schema, even though there is no actual schema data, just a tombstone. I've changed the logic to return no schema if the schema read back is a tombstone. The issue doesn't affect producers because the check was already correct there. I've also updated the check for transitive compatibility to remove the prefix of schemas before the deleted schema. Previously this was throwing an NPE on the broker as it couldn't decode the deleted schema. This issue was discovered by failures in the healthcheck. The check period (5 minutes) was longer than the GC period (60 seconds). I would expect it to hit quite often in other scenarios also.
jiazhai
pushed a commit
that referenced
this pull request
Aug 28, 2019
feca5bb changed topic delete logic to delete the schema when the topic is deleted (though this only seems to be enabled for idle topic GC). This exposed a bug in compatibility checking whereby if the a subscription tries to attach to the topic, even if using the same schema as had been used previously, a compatibility exception will be thrown. This is because the topic still appears to have a schema, even though there is no actual schema data, just a tombstone. I've changed the logic to return no schema if the schema read back is a tombstone. The issue doesn't affect producers because the check was already correct there. I've also updated the check for transitive compatibility to remove the prefix of schemas before the deleted schema. Previously this was throwing an NPE on the broker as it couldn't decode the deleted schema. This issue was discovered by failures in the healthcheck. The check period (5 minutes) was longer than the GC period (60 seconds). I would expect it to hit quite often in other scenarios also. (cherry picked from commit d77980d)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
feca5bb changed topic delete logic to delete the schema when the
topic is deleted (though this only seems to be enabled for idle topic
GC). This exposed a bug in compatibility checking whereby if the a
subscription tries to attach to the topic, even if using the same
schema as had been used previously, a compatibility exception will be
thrown.
This is because the topic still appears to have a schema, even though
there is no actual schema data, just a tombstone. I've changed the logic
to return no schema if the schema read back is a tombstone.
The issue doesn't affect producers because the check was already
correct there.
I've also updated the check for transitive compatibility to remove the
prefix of schemas before the deleted schema. Previously this was
throwing an NPE on the broker as it couldn't decode the deleted
schema.
This issue was discovered by failures in the healthcheck. The check
period (5 minutes) was longer than the GC period (60 seconds). I would
expect it to hit quite often in other scenarios also.