-
Notifications
You must be signed in to change notification settings - Fork 3.2k
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
BROKERFAIL : failed: err: Local: Bad message format: (errno: Resource temporarily unavailable) #1682
Comments
Huhm, it is weird that it doesn't say what response is malformed. It would also be valuable if you could reproduce this with latest librdkafka master |
@edenhill , Thanks .. Attaching logs with debug=all..
|
adding last 10K log line... |
So what I think is happening here is that the broker (or the network) is disconnecting the TCP session. |
Is there anything special with your setup? Load balancers? Firewalls? |
Let me enable Kafka Brokers DEBUG logs and will keep you posted here.... Just wondering first few minutes the connectivity seems to be working in other words we were able to produce/consume ~50K messages ... Is there anything special with your setup? Load balancers? Firewalls? - No.. Its my own test setup ... Thanks @edenhill for checking this ... |
@edenhill Yes , You are absolutely correct... Here is the DEBUG logs of Broker:
|
Broken pipe means something other than the broker tore down the connection, in your case that must be the client.
to:
and reproduce the issue with full debugging enabled? |
Thanks , I'll do it now .. Could you Pls provide steps to bundle from master ? How to bundle like this https://github.com/edenhill/librdkafka/archive/v0.11.3.tar.gz ?? Sorry this is the first time im trying LibRdKafkaDriver :-) .. |
The easiest approach is to:
|
Cool , Thanks .. I'm not directly running librdkafka program ... There is a another process which will use driver to run it .. So i need to give input file like this https://github.com/edenhill/librdkafka/archive/v0.11.3.tar.gz :-) ... |
You can create a tar ball of the change and provide that to your application perhaps. |
Cool , In few minutes , I'll attach logs ... |
@edenhill , Here you go :
|
Thanks! And this is with a patched librdkafka? |
Yep ..
|
@edenhill , Let me know if you are looking additional logs ... |
I can't find the printout that we patched, so it must be something else wrong. It could be that the returned FetchResponse is too big, but this should be logged appropriately. |
Client Settings: No luck after setting sink_err.txt Pls refer the attached logs again and let me know if you are able to see the expected printout ... |
Try using default settings (remove your config) for all of the max.bytes, fetch.bytes, etc. |
Sure , I'll run with recommended settings ... |
Still same issue ...
|
Attaching debug_all logs... |
@edenhill , Let me know if you want to try with different settings ... |
I'm a bit stumped to be honest, I can't find a location in the code where ERR__BAD_MSG is propagated without a debug message being emitted |
@edenhill , Did you find anything ?...Let me know if you are looking for additional debug logs etc ... also i could see consumer failed with same bad message error..
|
I want to add some extra debugging to track this down, would you be fine with testing a special branch? |
@edenhill - Yep , Im fine with testing a special branch .... |
@edenhill , Pls share me branch details if you are done with additional changes ... |
@edenhill , Ping ... |
Please try the |
Thanks , I'll test with this branch ... |
My current setup have some problem.. Will test this tomorrow... |
Sorry for the delay in testing ... without enabling debug_all i could see below:
|
Tried this : log.connection.close=true,receive.message.max.bytes=100000000
|
I've increased size to |
Consumer is running fine with load 5Mbps ... I'll increase the load and will keep you posted ... |
Any updates? |
After increasing receive.max.message.bytes I'm not finding bad error message. total no of partitions in topic : 30 ... |
@edenhill . I ran a test for three days continuously with load ~4Mbps and i dont see the this bad message error ... What is your recommendation of LibRdKafka Settings? Total Number of Partitions in a Topic : 30. Expected traffic ~30Mbps ...
I'll increase the load and let you know the status here.. Current Driver Settings:
|
That's a moderate traffic load and I think you can use default values for pretty much everything. |
@edenhill , I'm frequently getting message timeout error @7Mbps load.. No errors found in broker logs too.. Driver Settings: |
Enable debug=protocol and look at the response times (rtt) of requests, do they increase over time until finally timing out? You could try to set |
dup of #1616 |
Description
BROKERFAIL : failed: err: Local: Bad message format: (errno: Resource temporarily unavailable)
How to reproduce
Produce messages at rate 5MB/sec or 10MB/sec.
Checklist
Please provide the following information:
v0.11.3
Kafka version : 0.11.0.0 & Kafka commitId : cb8625948210849f
Ubuntu 16.04.2 LTS
debug=..
as necessary) from librdkafkaNo error message in broker
First , Pls check my client settings and let me know if anything should be modified...
Kafka Heap Settings:
echo $KAFKA_HEAP_OPTS
-Xms6g -Xmx20g -XX:MetaspaceSize=96m -XX:+UseG1GC -XX:MaxGCPauseMillis=20 -XX:InitiatingHeapOccupancyPercent=35 -XX:G1HeapRegionSize=16M -XX:MinMetaspaceFreeRatio=50 -XX:MaxMetaspaceFreeRatio=80
The text was updated successfully, but these errors were encountered: