could anyone take a look at this? <github.com/flue...
# fluent-bit
d
could anyone take a look at this? github.com/fluent/fluent-bit/issues/10821 -- no response from a dev since submitted Sept 2025
p
I've dropped a quick comment on it
d
thanks! I mentioned to the team to test on 4.2.
p
Note that 4.2 is EOM this month: github.com/fluent/fluent-bit/security/policy
l
It could be an issue with the HTTP/2 client code since it only happens when gRPC is enabled. It could also be that fluent-bit is not properly processing the gRPC response and thus not reporting the delivery failures to trigger a retry.
I haven't been involved in a long time but I saw this thread by chance and hopped on just because I worked on those components a few years ago.
I'd reccommend putting together a set of regular
fluent-bit -> fluent-bit
and
fluent-bit -> otel-collector
over
gRPC
configuration files to make it easier to reproduce and isolate the issue without extra cruft
You could also manually enable
http2
regardless of if you go through a "tradtional" regular HTTP/1.1 encapsulated
gRPC
packet in order to force fluent-bit to use the new HTTP client which will also try to use HTTP/2 if supported (although if you are going through a plaintext connection you'll have to set it to
force
in order to ensure that it doesn't fall back to HTTP/1.x)
That way you can get a bit closer to an answer (ie. determining if the problem is the new gen HTTP client dealing with the payload sizes, the gRPC implementation (at any stage) or whatever)
If you try to diagnose it on your own but find issues feel free to message me here so I get an alert and reply on a timely fashion since I don't keep tabs on the github repo