fix(snapshot): Replace sleeps with yield #5619
Open
+94
−29
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
dranikpg
commented
Aug 11, 2025
src/server/engine_shard.cc
Outdated
Comment on lines
541
to
432
| fb2::Fiber::Opts fb_opts{.priority = fb2::FiberPriority::BACKGROUND, .name = "heatbeat"}; | ||
| fiber_heartbeat_periodic_ = | ||
| MakeFiber([this, index = pb->GetPoolIndex(), period_ms, heartbeat]() mutable { | ||
| fb2::Fiber(fb_opts, [this, index = pb->GetPoolIndex(), period_ms, heartbeat]() mutable { |
There was a problem hiding this comment.
- Heartbeat should be background
- If its background, it doesn't penalize other background fibers on idle
- We can split it into stages to increase simplicity and have dynamic load on it
src/server/snapshot.cc
Outdated
Comment on lines
333
to
336
| // Yield after possibly long cpu slice due to compression and serialization | ||
| if (ThisFiber::Priority() == fb2::FiberPriority::BACKGROUND) | ||
| ThisFiber::Yield(); | ||
| // TODO: else Sleep() to provide write backpressure in advance? |
There was a problem hiding this comment.
Previously we yielded after the write, now we yield before because that's the end of the "CPU" slice (that the scheduler masures). There shoudn't be any difference as long as writes are not async (not sure why we didn't add that yet) 🤔
dranikpg
commented
Aug 11, 2025
| @@ -164,8 +164,11 @@ class SliceSnapshot : public journal::JournalConsumerInterface { | |||
|
|
|||
| // Used for sanity checks. | |||
| bool serialize_bucket_running_ = false; | |||
|
|
|||
| std::string snapshot_fb_name_; | |||
There was a problem hiding this comment.
helio fiber stores a string view, previously it pointed just to a stack placed string 😳
|
Will fix heartbeat after romange/helio#449 |
dranikpg
force-pushed
the
replication-yields
branch
2 times, most recently
from
6925b6d to
e3946d5
Compare
|
|
I didn't manage to find a "winning" strategy for finding the best scheduler configuration, so I just resorted to finding the best default value 😞 Parameters for one of cloud's snapshots 
|
|
Please rebase. Is it ready for review @dranikpg ? |
Signed-off-by: Vladislav Oleshko <vlad@dragonflydb.io>
Signed-off-by: Vladislav Oleshko <vlad@dragonflydb.io>
Signed-off-by: Vladislav Oleshko <vlad@dragonflydb.io>
dranikpg
force-pushed
the
replication-yields
branch
from
e3946d5 to
c96b31f
Compare
Signed-off-by: Vladislav Oleshko <vlad@dragonflydb.io>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Blocked on new background fibers from helio romange/helio#402.
Background tasks should be as cooperative as possible to not increase the latency. We yield after each bucket traversal either by pushing to the channel or by yielding explicitly