Documentation
¶
Overview ¶
Package idspace partitions the virtual user id range across a cluster.
Every virtual user in a run needs an id that is unique fleet-wide: ids seed each VU's random source, pick its slice of shared fixtures, and identify it in logs. Asking a central allocator for one would put a network round trip in the middle of scaling up. Instead the range is carved statically — the coordinator gives each agent a block, the agent subdivides it for its workers — so allocation is a local increment and collisions are impossible by construction.
Index ¶
Constants ¶
const ( // PerWorker is how many ids each worker process may allocate. PerWorker int64 = 100_000 // PerAgent is how many ids each agent may hand out across its workers, // allowing up to a hundred workers per agent. PerAgent int64 = PerWorker * 100 )
Block sizes. Generous, because the space is 63 bits wide and the only cost of a large block is that ids are not densely packed: at these strides a run can span roughly nine hundred thousand agents before it exhausts the range, which is comfortably more than anyone will need.
Variables ¶
This section is empty.
Functions ¶
func WorkerBase ¶
WorkerBase returns the first id belonging to one worker of an agent.
A worker index beyond the block wraps rather than spilling into the next agent's range. Overlapping ids inside one agent would be a nuisance; ids colliding across hosts would corrupt data sharding, so the wrap keeps the damage local and bounded.
Types ¶
This section is empty.
Source Files