What "no cap" covers
On a 24/7 plan you raise as many production incidents as you need to, and every one gets a root cause analysis and a post-incident review. These are the Flink problems we work on most:
Flink CDC jobs that fall behind on large batches, miss deletes from the source table, or cause deadlocks in the source database
JobManager and TaskManager crash loops, including TaskManagers running out of Metaspace
Jobs that fail on submission, or stop after a storage or authentication change
Autoscaling into an OutOfMemoryError when a traffic spike hits
Two jobs with the same configuration and the same input performing very differently
Lookup joins that overload the system they look up against
Events reaching the sink out of order
What's not included is new work: architecture and design, proactive performance projects, capacity planning, migrations, new features and extensions, and training. That work is billed as support hours. We classify every ticket as an incident or a consulting request and tell you which, so you always know what draws from your hours.
Our engineers are senior Apache Flink specialists who work on complex production problems. We don't run a first-line help desk.