Bring back the old behavior
Signed-off-by: Sasha Bogicevic <[email protected]>
Home /
Input Output /
hydra-poc
Jan 22, 10-11 AM (0)
Jan 22, 11-12 PM (0)
Jan 22, 12-1 PM (0)
Jan 22, 1-2 PM (9)
Jan 22, 2-3 PM (1)
Jan 22, 3-4 PM (0)
Jan 22, 4-5 PM (2)
Jan 22, 5-6 PM (0)
Jan 22, 6-7 PM (0)
Jan 22, 7-8 PM (7)
Jan 22, 8-9 PM (0)
Jan 22, 9-10 PM (0)
Jan 22, 10-11 PM (0)
Jan 22, 11-12 AM (0)
Jan 23, 12-1 AM (0)
Jan 23, 1-2 AM (0)
Jan 23, 2-3 AM (0)
Jan 23, 3-4 AM (0)
Jan 23, 4-5 AM (0)
Jan 23, 5-6 AM (1)
Jan 23, 6-7 AM (0)
Jan 23, 7-8 AM (0)
Jan 23, 8-9 AM (1)
Jan 23, 9-10 AM (0)
Jan 23, 10-11 AM (1)
Jan 23, 11-12 PM (1)
Jan 23, 12-1 PM (0)
Jan 23, 1-2 PM (1)
Jan 23, 2-3 PM (2)
Jan 23, 3-4 PM (0)
Jan 23, 4-5 PM (1)
Jan 23, 5-6 PM (0)
Jan 23, 6-7 PM (0)
Jan 23, 7-8 PM (0)
Jan 23, 8-9 PM (0)
Jan 23, 9-10 PM (0)
Jan 23, 10-11 PM (0)
Jan 23, 11-12 AM (0)
Jan 24, 12-1 AM (0)
Jan 24, 1-2 AM (0)
Jan 24, 2-3 AM (0)
Jan 24, 3-4 AM (0)
Jan 24, 4-5 AM (0)
Jan 24, 5-6 AM (0)
Jan 24, 6-7 AM (0)
Jan 24, 7-8 AM (0)
Jan 24, 8-9 AM (0)
Jan 24, 9-10 AM (0)
Jan 24, 10-11 AM (0)
Jan 24, 11-12 PM (0)
Jan 24, 12-1 PM (0)
Jan 24, 1-2 PM (0)
Jan 24, 2-3 PM (0)
Jan 24, 3-4 PM (0)
Jan 24, 4-5 PM (0)
Jan 24, 5-6 PM (0)
Jan 24, 6-7 PM (0)
Jan 24, 7-8 PM (0)
Jan 24, 8-9 PM (0)
Jan 24, 9-10 PM (0)
Jan 24, 10-11 PM (0)
Jan 24, 11-12 AM (0)
Jan 25, 12-1 AM (0)
Jan 25, 1-2 AM (0)
Jan 25, 2-3 AM (0)
Jan 25, 3-4 AM (0)
Jan 25, 4-5 AM (0)
Jan 25, 5-6 AM (0)
Jan 25, 6-7 AM (0)
Jan 25, 7-8 AM (0)
Jan 25, 8-9 AM (0)
Jan 25, 9-10 AM (0)
Jan 25, 10-11 AM (0)
Jan 25, 11-12 PM (0)
Jan 25, 12-1 PM (2)
Jan 25, 1-2 PM (0)
Jan 25, 2-3 PM (1)
Jan 25, 3-4 PM (0)
Jan 25, 4-5 PM (0)
Jan 25, 5-6 PM (0)
Jan 25, 6-7 PM (0)
Jan 25, 7-8 PM (0)
Jan 25, 8-9 PM (0)
Jan 25, 9-10 PM (0)
Jan 25, 10-11 PM (0)
Jan 25, 11-12 AM (0)
Jan 26, 12-1 AM (0)
Jan 26, 1-2 AM (0)
Jan 26, 2-3 AM (0)
Jan 26, 3-4 AM (0)
Jan 26, 4-5 AM (0)
Jan 26, 5-6 AM (0)
Jan 26, 6-7 AM (0)
Jan 26, 7-8 AM (0)
Jan 26, 8-9 AM (0)
Jan 26, 9-10 AM (0)
Jan 26, 10-11 AM (0)
Jan 26, 11-12 PM (0)
Jan 26, 12-1 PM (2)
Jan 26, 1-2 PM (1)
Jan 26, 2-3 PM (2)
Jan 26, 3-4 PM (1)
Jan 26, 4-5 PM (1)
Jan 26, 5-6 PM (1)
Jan 26, 6-7 PM (0)
Jan 26, 7-8 PM (0)
Jan 26, 8-9 PM (0)
Jan 26, 9-10 PM (0)
Jan 26, 10-11 PM (0)
Jan 26, 11-12 AM (0)
Jan 27, 12-1 AM (0)
Jan 27, 1-2 AM (0)
Jan 27, 2-3 AM (0)
Jan 27, 3-4 AM (0)
Jan 27, 4-5 AM (0)
Jan 27, 5-6 AM (0)
Jan 27, 6-7 AM (0)
Jan 27, 7-8 AM (9)
Jan 27, 8-9 AM (2)
Jan 27, 9-10 AM (1)
Jan 27, 10-11 AM (1)
Jan 27, 11-12 PM (11)
Jan 27, 12-1 PM (11)
Jan 27, 1-2 PM (1)
Jan 27, 2-3 PM (0)
Jan 27, 3-4 PM (1)
Jan 27, 4-5 PM (1)
Jan 27, 5-6 PM (1)
Jan 27, 6-7 PM (1)
Jan 27, 7-8 PM (2)
Jan 27, 8-9 PM (0)
Jan 27, 9-10 PM (0)
Jan 27, 10-11 PM (0)
Jan 27, 11-12 AM (0)
Jan 28, 12-1 AM (0)
Jan 28, 1-2 AM (0)
Jan 28, 2-3 AM (0)
Jan 28, 3-4 AM (0)
Jan 28, 4-5 AM (0)
Jan 28, 5-6 AM (0)
Jan 28, 6-7 AM (0)
Jan 28, 7-8 AM (0)
Jan 28, 8-9 AM (4)
Jan 28, 9-10 AM (2)
Jan 28, 10-11 AM (13)
Jan 28, 11-12 PM (2)
Jan 28, 12-1 PM (1)
Jan 28, 1-2 PM (0)
Jan 28, 2-3 PM (0)
Jan 28, 3-4 PM (1)
Jan 28, 4-5 PM (0)
Jan 28, 5-6 PM (0)
Jan 28, 6-7 PM (0)
Jan 28, 7-8 PM (0)
Jan 28, 8-9 PM (0)
Jan 28, 9-10 PM (0)
Jan 28, 10-11 PM (0)
Jan 28, 11-12 AM (10)
Jan 29, 12-1 AM (0)
Jan 29, 1-2 AM (0)
Jan 29, 2-3 AM (0)
Jan 29, 3-4 AM (0)
Jan 29, 4-5 AM (0)
Jan 29, 5-6 AM (0)
Jan 29, 6-7 AM (0)
Jan 29, 7-8 AM (0)
Jan 29, 8-9 AM (11)
Jan 29, 9-10 AM (0)
Jan 29, 10-11 AM (0)
124 commits this week
Jan 22, 2026
-
Jan 29, 2026
Fix ci for running benchmarks
Not sure if reusing scaling-factor is good here Signed-off-by: Sasha Bogicevic <[email protected]>
Make dataset command only generate datasets
Signed-off-by: Sasha Bogicevic <[email protected]>
Fix CI regarding the latest changes in running the e2e benchmark
Signed-off-by: Sasha Bogicevic <[email protected]>
Move dataset generation command
When generating datasets we also run benchmarks but using a `single` command just runs the existing dataset. Signed-off-by: Sasha Bogicevic <[email protected]>
Rename single to standalone since the name is misleading
Signed-off-by: Sasha Bogicevic <[email protected]>
Re-introduce snapshot in PartySignedSnapshot
We should remove this in some regular pr not dealing with benchmarks only Signed-off-by: Sasha Bogicevic <[email protected]>
Introduce bench type: constant or growing
Signed-off-by: Sasha Bogicevic <[email protected]>
Draft a growing utxo benchmark dataset
Introduce number of fanout outputs and number of txs
Now we pass in wanted number of txs and we also output the number of fanout outputs so we can track the fanout performance. Signed-off-by: Sasha Bogicevic <[email protected]>
WIP: still cannot get reliable reproduction
Signed-off-by: Sasha Bogicevic <[email protected]>
introduce new chainPointTime
> NodeState now tracks currentChainPoint :: ChainPointType tx, instead of currentSlot :: ChainSlot + currentChainTime :: Maybe UTCTime > NodeSynced and NodeUnsynced now carry ChainPointType tx, instead of chainTime :: UTCTime > extended IsChainState class with chainPointTime :: ChainPointType tx -> UTCTime > initialChainTime = posixSecondsToUTCTime 0
request for changes
* extended NodeSynced and NodeUnsynced with time and slot drifts * added Haddocks to both currentSlot and currentChainTime * made NodeState.currentChainTime non optional * defined initialChainTime = posixSecondsToUTCTime 0
Report drift on "how out of sync" the node is
do not track drift measured in slot
> because we have no recent view of the chain since we are out of sync
fix arithmetic underflow
> when calculating the slot drift
use latest NodeState.current slot instead of latest chain point tracked
> the former gets updated upon ticks the later upon head transitions observed, which usually is older than ticks
Revert "introduce new chainPointTime"
This reverts commit 037aea1d9f15f65db00113f0606f9e09d0b03548.
enhance tui report on node out of sync
WIP: still cannot get reliable reproduction
Signed-off-by: Sasha Bogicevic <[email protected]>
Revert input and etcd bounded queues sizes back to 100 (#2466)
<!-- Describe your change here --> Follow-up https://github.com/cardano-scaling/hydra/pull/2430 > Reduce the size of the input and etcd pending broadcast queues so they are smaller than the logging queue. When running `cabal run bench-e2e -- single --cluster-size 3 --scaling-factor 100`, having upstream queues as large as or larger than the logging queue led to invalid transactions under load. This change enforces proper back pressure and prevents that behavior. --- <!-- Consider each and tick it off one way or the other --> * [X] CHANGELOG updated or not needed * [X] Documentation updated or not needed * [X] Haddocks updated or not needed * [X] No new TODOs introduced or explained herafter
revert input and etcd bounded queues sizes back to 100
> they need to be smaller than the logging queue size which is 500 by default