Repository, JVM, operating system and queue settings that decide NiFi throughput, plus how to grow and shrink clusters, based on the NiFi Administrator's and User Guides.
Planning disks for the FlowFile, content, provenance and status repositories: separate volumes, multiple content and provenance directories, sizing from data rate and retention, and the failure mode when repositories share a disk.
Choosing provenance retention time and size limits, index shard size, multiple storage directories, and when the volatile or no-op provenance repositories make sense for high-volume flows that do not need lineage.
How content archiving affects disk usage and provenance replay: retention period, archive usage percentage, the back pressure percentage that blocks writes, and settings for flows that move large files.
Setting heap size and garbage collection for NiFi in bootstrap.conf, why large attributes and big queues pressure the heap, code cache tuning under sustained throughput and heap dumps on out-of-memory errors.
Setting object and data size thresholds per connection and per process group, FlowFile expiration with prioritizers, and how thresholds interact with swapping, content repository growth and slow destinations.
Spreading data across cluster nodes after primary-node sources with round robin, partition by attribute or single node strategies: compression, behavior when a node disconnects and the cluster connection summary.
The operating system settings NiFi documents for IO-heavy flows: file handle and process limits, TCP port range and TIME_WAIT, swappiness, noatime mounts and antivirus exclusions for repository directories.
What happens when queues grow past the swap threshold: heap protection, ordering caveats with prioritizers, the effect on throughput and flow designs that keep queues small.
Turning on processor performance tracking to see CPU time, content repository reads and writes and garbage collection time per processor, the sampling cost, and reading status history to find the slow step.
Scaling a NiFi cluster up and down safely: joining nodes with the current flow, disconnect, offload and delete steps, persisted node lists that block redistribution until nodes reconnect, and doing it on Kubernetes.
Why list processors such as ListSFTP, ListSmb and ListGoogleDrive run on the primary node only, how state carries over when the primary changes, and the flow analysis rule that requires a load-balanced connection after them.
Tell us what you're working on, whether it's a deployment, an audit, a security test or a cyber range. You'll speak with an engineer who can help you scope it.