jepsen.history.sim.gen — io.jepsen/history.sim 0.1.1

A minimal recreation of generators for simulated DB ops. We're not doing anything fancy with context or causality here; generators are just plain old seqs.

These versions use clojure.data.gen for RNGs, so they're deterministic.

append-txns clj

(append-txns opts)

Like wr-txns, we just rewrite writes to be appends.

fresh-key clj

(fresh-key active-keys k)

Takes a key and a vector of active keys. Returns the vector with that key replaced by a fresh key.

generators clj

A map of generator names to constructor fns.

key-dist-scale clj

(key-dist-scale key-dist-base key-count)

Takes a key-dist-base and a key count. Computes the scale factor used for random number selection used in rand-key.

list-append clj

(list-append)

(list-append opts)

A generator for operations where values are transactions made up of reads and appends to various integer keys. Takes options:

:key-count Number of distinct keys at any point :min-txn-length Minimum number of operations per txn :max-txn-length Maximum number of operations per txn :max-writes-per-key Maximum number of operations per key

For defaults, see wr-txns.

rand-key clj

(rand-key key-dist key-dist-base key-dist-scale active-keys)

Helper for generators. Takes a key distribution (e.g. :uniform or :exponential), a key distribution scale, a key distribution base, and a vector of active keys. Returns a random active key.

txn-gen clj

(txn-gen txns)

Takes a sequence of transactions and returns a sequence of invocation operations.

wr-txns clj

(wr-txns opts)

(wr-txns opts state)

A lazy sequence of write and read transactions over a pool of n numeric keys; every write is unique per key. Options:

:key-dist Controls probability distribution for keys being selected for a given operation. Choosing :uniform means every key has an equal probability of appearing. :exponential means that key i in the current key pool is k^i times more likely than the first key to be chosen. Defaults to :exponential.

:key-dist-base The base for an exponential distribution. Defaults to 2, so the first key is twice as likely as the second, which is twice as likely as the third, etc.

:key-count Number of distinct keys at any point. Defaults to 10 for exponential, 3 for uniform.

:min-txn-length Minimum number of operations per txn

:max-txn-length Maximum number of operations per txn

:max-writes-per-key Maximum number of operations per key