-
Type: Bug
-
Status: Resolved
-
Priority: Major
-
Resolution: Duplicate
-
Affects Version/s: 10.10
-
Fix Version/s: None
-
Component/s: Streams
-
Tags:
-
Backlog priority:500
-
Team:PLATFORM
-
Sprint:nxplatform #19, nxplatform #20, nxplatform #21, nxplatform #22, nxplatform #29
-
Story Points:3
When a non-default codec is configured for the audit,
for instance in 10.10 we add to our nuxeo.conf:
nuxeo.stream.audit.log.codec=avro
while the default is in 10.10 is legacy.
After a while, you restart a node but you have an error, for instance Kafka is not reachable and you are not using the strict mode (relax is on), resulting in a Nuxeo started with the stream service down:
= Component Loading Status: Pending: 0 / Missing: 0 / Unstarted: 1 / Total: 670 service:org.nuxeo.runtime.stream.service
If Kafka is now reachable It seems possible that this instance is able to produce audit records in the audit stream, but it is going to use the default codec instead of the custom one.
The result is that the audit stream processing that saves the log entries to the audit backend will fail systematically because of the bad codec.
This needs to be reproduce in order to make sure that:
- using strict mode will prevent this
- the kafka checker added in HF32 is able to avoid causing the problem