Feature Request: rate limit compaction but not flushes about rocksdb

[ajkr]

Oh I see. A simpler thing to expose in the C API would be the mode argument:

rocksdb/include/rocksdb/rate_limiter.h

Line 165 in 4835c11

RateLimiter::Mode mode = RateLimiter::Mode::kWritesOnly,

Then rate limit RateLimiter::Mode::kReadsOnly, which would have the side effect of not rate limiting flush because it does not perform read I/O

[ajkr]

From that wiki - "RocksDB specified rocksdb::Env::IO_LOW for internal compaction related IO and rocksdb::Env::IO_HIGH for internal flush related IO."

To achieve your goal, you could implement a RateLimiter that wraps a rate limiter created by NewGenericRateLimiter(). For requests with priority below Env::IO_HIGH, forward them to the wrapped rate limiter. For other requests, do not forward them and instead simply return.

[zaidoon1]

so for more context, I'm using rust-rocksdb which relies on the c api to use rocksdb. Correct me if i'm wrong, but it looks like the best (only?) way to do this would be to update the rate limiter in rocksdb to allow more customization/decide what to rate limit and update the c api accordingly?

[ajkr]

Fixed by #12259