-
Notifications
You must be signed in to change notification settings - Fork 642
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
Importer(s) are not communicating import done correctly #26046
Comments
## Description The importers were not correctly marking their import position documents as completed to signify the record readers were done. This was due to missing zeebe indices causing a `NoSuchIndex` exception which means the path through which record reader completion was checked and set was missed. The fix involves treating a no such index exception as an empty batch with regards to record reader completion check as they signify the same thing (no records to export). ## Checklist - [x] Given missing zeebe indices record readers can still get marked as completed. ## Related issues closes #26046
@ChrisKujawa I can still reproduce it with 8.7.0-alpha3. I'm using docker-compose with CamundaExporter for Zeebe. |
Can you give me the exact steps to reproduce? Is it consistent or non-deterministic? |
Hi @EuroLew, I can reliably reproduce it. I used docker-compose to start alpha3 with CamundaExporter. The issue occurs every time I restart the whole docker compose. |
@EuroLew please take a look at this, this week. As discussed here https://camunda.slack.com/archives/C06F0GLJNFM/p1738926629778839?thread_ts=1738775414.411889&cid=C06F0GLJNFM
|
will do |
## Description There is a bug where the camunda exporter can enter an error loop in a new cluster where it is waiting for importers to complete which never will (as importers will not do any work in 8.8). This fix will as per the slack thread referred to here #26046 (comment), assume that a lack of existence of zeebe indices (which the importer reads from), means that it is a fresh cluster, thus the exporter does not need to wait and we avoid the error loop. Questions - There is a requirement that the new `zeebeIndexPrefix` match the zeebe index prefix defined for the ES exporter, I have documented this in the yaml but I feel like this should be explicitly stated elsewhere otherwise the 8.8 camunda exporter can enter boot loops. (Although it will work work with default configuration). - Would this need to get backported for people running benchmarks? (to which version?) ## Related issues relates to #26046
Should be resolved with #28105 |
Actually we need to validate the version of the Zeebe indices, like 8.7. IF 8.8 only is in the zeebe indices, then it is still fine to assume greenfield. |
Will be covered with #28833 |
Describe the bug
When running a benchmark, we observed that the Camunda Exporter failed after restart, because the Importer doesn't seem to be completed (where it should be).
The importer seems to not complete all documents.
The text was updated successfully, but these errors were encountered: