-
Notifications
You must be signed in to change notification settings - Fork 200
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
[Bug]: Consuming right after eviction results in No such file or directory
#3891
Comments
Added a repro link. |
No such file or directory
No such file or directory
No such file or directory
I was mistaken in the fact it happens just with an absolute offset - the same thing occurs with |
@SecretPocketCat thanks for reporting. |
Looks like there are two problems here.
@SecretPocketCat After all these fixes, you will still need to manually restart the consuming. We can't do it implicitly, because this is the kind of decision that every user has to make according to project specifics. |
@SecretPocketCat can you please confirm if the fix worked? Please use latest to check: fvm install latest To swith back to the production version, please use: fvm switch stable |
@ajhunyady It does work. |
My assumption is that this was done as a convenience to the client to avoid an error, though it does seem to defeat the purpose of the error. @galibey can you please shed some light on this? |
@SecretPocketCat In your repro example there is no eviction occurring there. The ...
spec.set_storage(TopicStorageConfig {
segment_size: Some(10000),
max_partition_size: Some(20000),
});
...
let producer = fluvio
.topic_producer_with_config(
topic_name.clone(),
TopicProducerConfigBuilder::default()
.batch_size(9000)
.build()?,
)
.await?;
... Also, make sure you use the latest |
@galibey Alright, now it makes sense. Thank you. |
FYI you can also do |
What happened
Trying to consume right after records have been evicted results in an FS error (I assume that's for the segment containing that record).
The fetched offsets used for resolving logical offsets are stale.
Expected behavior
The consumer will use continue from the first valid/non-evicted offset
Describe the setup
How to reproduce it (as minimally and precisely as possible)
repro repo
Steps to reproduce the behavior:
Additional context
I'm using an absolute offset to track offsets (before fluvio offset management lands).
The text was updated successfully, but these errors were encountered: