A change included in
https://github.com/solana-labs/solana/pull/20480
was that when the root node in turbine broadcast tree is down, the
leader will broadcast the shred to all nodes in the first layer.
The intention was to mitigate the impact of dead nodes on shreds
propagation, because if the root node is down, then the entire cluster
will miss out the shred.
On the other hand, if x% of stake is down, this will cause 200*x% + 1
packets/shreds ratio at the broadcast stage which might contribute to
line-rate saturation and packet drop.
To avoid this bandwidth saturation issue, this commit reverts that logic
and always broadcasts shreds from the leader only to the root node.
As before we rely on erasure codes to recover shreds lost due to staked
nodes being offline.
Fully deserializing shreds in window-service before sending them to
retransmit stage adds latency to shreds propagation.
This commit instead channels through the payload and relies on only
partial deserialization of a few required fields: slot, shred-index,
shred-type.
In prepration of
https://github.com/solana-labs/solana/pull/25807
which reworks erasure batch sizes, this commit:
* adds a helper function mapping the number of data shreds to the
erasure batch size.
* adds ProcessShredsStats to Shredder::entries_to_shreds in order to
replace and remove entries_to_data_shreds from the public interface.
The indices for erasure coding shreds are tied to data shreds:
https://github.com/solana-labs/solana/blob/90f41fd9b/ledger/src/shred.rs#L921
However with the upcoming changes to erasure schema, there will be more
erasure coding shreds than data shreds and we can no longer infer coding
shreds indices from data shreds.
The commit adds constructs to track coding shreds indices explicitly.
next-shred-index is already readily available from returned data shreds.
The commit simplifies the api for upcoming changes to erasure coding
schema which will require explicit tracking of indices for coding shreds
as well as data shreds.
Turbine randomly shuffles cluster nodes on a broadcast tree for each
shred. This requires knowing the stakes and nodes' contact-infos (from
gossip).
However gossip is subject to partitioning and propogation delays.
Additionally unstaked nodes may join and leave the cluster at any
moment, changing the cluster view from one node to another.
This commit:
* Always arranges the unstaked nodes at the bottom of turbine broadcast
tree.
* Staked nodes are always included regardless of if their contact-info
is available in gossip or not.
* Uses the unbiased WeightedShuffle construct for shuffling nodes.
bank.get_leader_schedule_epoch(shred_slot)
is one epoch after epoch_schedule.get_epoch(shred_slot).
At epoch boundaries, shred is already one epoch after the root-slot. So
we need epoch-stakes 2 epochs ahead of the root. But the root bank only
has epoch-stakes for one epoch ahead, and as a result looking up epoch
staked-nodes from the root-bank fails.
To be backward compatible with the current master code, this commit
implements a fallback on working-bank if epoch staked-nodes obtained
from the root-bank is none.
* Current caching mechanism does not update cluster-nodes when the epoch
(and so epoch staked nodes) changes:
https://github.com/solana-labs/solana/blob/19bd30262/core/src/broadcast_stage/standard_broadcast_run.rs#L332-L344
* Additionally, the cache update has a concurrency bug in which the
thread which does compare_and_swap may be blocked when it tries to
obtain the write-lock on cache, while other threads will keep running
ahead with the outdated cache (since the atomic timestamp is already
updated).
In the new ClusterNodesCache, entries are keyed by epoch, and so if
epoch changes cluster-nodes will be recalculated. The time-to-live
eviction policy is also encapsulated and rigidly enforced.
carllin
behzad nouri
Michael Vines
Ikko Ashimine
Justin Starry
sakridge
jbiseda
Alexander Meißner