Roman Gershman
Roman Gershman
@adiholden have you tried using old (rdb) format?
It's indeed a behaviour difference. @safa-topal is it possible to workaround this issue on the client side by creating an async task that calls "BGSAVE"?
also IORING_SETUP_SINGLE_ISSUER IORING_SETUP_TASKRUN_FLAG
Consider checking with Sidekiq - https://github.com/search?q=repo%3Asidekiq%2Fsidekiq%20redis_version&type=code The maintainer of the framework may even agree to change the code to adapt to both standards but it will still be a problem...
Also https://github.com/StackExchange/StackExchange.Redis/blob/2f69707d87488febb7376739ed4dc51ad9a422c6/src/StackExchange.Redis/ResultProcessor.cs#L864
@manast Thank you for finding this! We are investigating whether it's on Dragonfly side.
@ddevault but nobody asked you to die. Why suddenly such passion? Have you contributed to the project before? This is how it appears to an outsider like me: While Redis'...
wanna send a PR?
@axboe, one of Dragonfly's users, also reported this as a behavioural change: Dragonfly, which does not use disk IO, bumps up the IOWAIT metric to 100%. If they run it...
A great discussion about this topic on [lore.kernel.org](https://lore.kernel.org/all/[email protected]/T/#mc2fa25c51f5ff42a8b35de95a8ce6f3c40b23d33). Just for us to understand, once we call `io_uring_register_iowait`, it will flag networking I/O as iowait but iouring will run in more...