Modifiable Broker Settings

Instaclustr Apache Kafka clusters are configured with optimal settings for our managed service. If required, the Kafka broker configuration items can be changed by submitting a support request, following are a few common configurations:

Name  Description  Type  Default 
auto.create.topics.enable  Enable automatic creation of a topic when a client produces to or consumes from a topic that does not exist  boolean  true 
delete.topic.enable  Allow topics to be deleted. If this is off, a delete request has no effect  boolean  true 
default.replication.factor  Replication factor used for topics that are created automatically  int  3 
num.partitions  Number of partitions used for topics that are created automatically  int  3 
compression.type  Specify the final compression type for a given topic. This configuration accepts the standard compression codecs (‘gzip’, ‘snappy’, ‘lz4’). It additionally accepts ‘uncompressed’ which is equivalent to no compression; and ‘producer’ which means retain the original compression codec set by the producer.  string  producer 
connections.max.idle.ms  How long a client connection can stay idle before the broker closes it  int  600000 
group.initial.rebalance.delay.ms  How long the group coordinator waits for more consumers to join a new group before the first rebalance  int  3000 
jute.maxbuffer  Maximum size, in bytes, of data stored in a ZooKeeper znode. Applies to ZooKeeper-based clusters  int  null 
leader.imbalance.check.interval.seconds  How often the partition leader rebalance check runs  int  600 
log.cleaner.threads  Number of background threads used for log cleaning  int  1 
log.retention.bytes *  Maximum size of a log before deleting it, in bytes. -1 means no size limit. This is the default for topics that do not set their own retention  long  -1 
log.retention.ms *  How long to keep a log before deleting it, in milliseconds. -1 means no time limit. This is the default for topics that do not set their own retention  long  null 
message.max.bytes  Maximum size of a message the broker will accept  int  1048588 
min.insync.replicas  Minimum number of in-sync replicas required for a write with acks=all to succeed  int  1 
num.network.threads  The number of threads that the server uses for receiving requests from the network and sending responses to the network  int  3 
offsets.retention.minutes  How long committed offsets are kept after a consumer group becomes inactive  int  10080 
remote.log.reader.max.pending.tasks  Maximum number of queued remote-log read tasks. Applies when Tiered Storage is enabled  int  100 
remote.log.reader.threads  Thread pool size for remote-log reads. Applies when Tiered Storage is enabled  int  10 
replica.fetch.max.bytes  Maximum bytes a follower fetches from the leader in one request  int  1048576 
replica.selector.class  Class that chooses which replica serves a consumer fetch. Set this to enable follower fetching  string  null 
replica.socket.receive.buffer.bytes  Receive buffer for sockets used when replicating from the leader. -1 uses the operating system default  int  65536 
socket.listen.backlog.size  Maximum number of pending connections on a broker socket  int  50 
socket.receive.buffer.bytes  Receive buffer for broker sockets. -1 uses the operating system default  int  102400 
socket.send.buffer.bytes  Send buffer for broker sockets. -1 uses the operating system default  int  102400 
transaction.state.log.min.isr  Minimum number of replicas that must acknowledge a write to the transaction topic  int  2 
unclean.leader.election.enable  Allow a replica that is not in the in-sync replica set to be elected leader. This can lose data  boolean  false 

* These log settings are only applicable for topics that are created via auto topic creation. These settings can be specified on a per topic basis when the topic is created explicitly.

The first four are chosen in the cluster creation form and written into server.properties from that saved choice. leader.imbalance.check.interval.seconds is 600 on current clusters. Kafka’s own default is 300, and that is what clusters on Kafka 3.5.0 and earlier still use.