[CHIA-1638] Pace block requests #18729
Merged
+69
−5
Conversation
This file contains 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
arvidn
force-pushed
the
pace-block-requests
branch
from
d5768f2
to
432e953
Compare
arvidn
added
the
Changed
|
This pull request has conflicts, please resolve those before we can evaluate the pull request. |
github-actions
bot
added
the
merge_conflict
arvidn
force-pushed
the
pace-block-requests
branch
from
432e953
to
ff3ffd6
Compare
github-actions
bot
removed
the
merge_conflict
|
Conflicts have been resolved. A maintainer will review the pull request shortly. |
arvidn
force-pushed
the
pace-block-requests
branch
from
ff3ffd6
to
ac7893a
Compare
github-actions
bot
added
the
merge_conflict
|
This pull request has conflicts, please resolve those before we can evaluate the pull request. |
arvidn
force-pushed
the
pace-block-requests
branch
from
ac7893a
to
ced7e0f
Compare
|
Conflicts have been resolved. A maintainer will review the pull request shortly. |
github-actions
bot
removed
the
merge_conflict
almogdepaz
reviewed
Oct 30, 2024
emlowe
reviewed
Oct 30, 2024
emlowe
reviewed
Oct 30, 2024
altendky
previously requested changes
Oct 31, 2024
emlowe
previously approved these changes
Nov 1, 2024
arvidn
added
ready_to_merge
github-actions
bot
added
the
merge_conflict
|
This pull request has conflicts, please resolve those before we can evaluate the pull request. |
arvidn
force-pushed
the
pace-block-requests
branch
from
d451773
to
6dc9af2
Compare
|
Conflicts have been resolved. A maintainer will review the pull request shortly. |
github-actions
bot
removed
the
merge_conflict
emlowe
approved these changes
Nov 1, 2024
|
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Outdated suggestions cannot be applied.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.
Purpose:
This patch addresses a problem during long-sync, where the syncing node may send
request_blocksat a rate exceeding the peer's outbound rate limit forrespond_blocks. The result of which is the peer not responding and the syncing node closing the connection. Over time, all peers may get disconnected and syncing stalls. It takes at least 30 seconds + weight proof validation to restart syncing.The root of this problem is that the rate limits for
request_blocksis not aligned with the rate limit forrespond_blocks. They are 500 msgs/minute and 100 msgs/minute respectively (but then scaled to 30%). It never makes sense to send more requests that the peer is willing to respond to, so these limits should really be the same.This patch hard-codes the expected outbound rate limit of the peer and paces requests to never exceed that limit. Thus, maintaining a steady sync (albeit, slow). The rate is at most one request every 2 seconds.
We already send requests to multiple peers, if we have more than one. This patch keeps track of the timestamp, per peer, when it's OK to send the next request. Sometimes, this timestamp can be in the past. This happens if one peer stalls for a long time and we "miss" the time to send a request to another peer (or the same peer for that matter).
The rate limit is enforced at 60 seconds at a time, so we allow "catching up" by only incrementing the timestamp by the rate limit minimum (2 seconds). However, if a peer takes too long to respond, we penalize it by bumping the time stamp to the current time. This creates a weak affinity to request more from faster peers.
There are still issues with our concept of rate limits. For instance, there is a configuration option to scale the rate limits. But the effective limits are never communicated over the protocol, so there's no way of knowing whether a peer has tweaked its limits.
Current Behavior:
During long sync, we request blocks as fast as we can (with a single request outstanding at a time). Risking pushing peers over the limit, stalling and having to restart the sync.
New Behavior:
During long sync, we pace the block requests to peers to never exceed the (presumed) rate limit for block requests.
Testing Notes:
Manually tested on my node. I sync about 3x faster.