Fix benchmarks

[appetrosyan]

MWE

run

cargo bench

Expected

Works

Actual

Error messages in logs complaining about not being able to find alice@1@wonderland.

[s8sato]

Disambiguation. If this issue refers to WARN logs such as below on running cargo bench --bench tps-dev, then #2078 is responsible for it. This occurs when the interval_us_per_tx configuration is too short.

Apart from that, I found another issue by accident. The tps-dev benchmark seems to have broken at some point since #1963. As of efa9a35692a949707ae0752119e4c4a1a08342a1, it fails with

thread 'main' panicked at 'The block is not yet in WSV. Need more sleep?', client/benches/tps/lib.rs:94:22

message, which would not be resolved by increasing the sleep (at least in 32 seconds, I've confirmed). This means the same number of blocks as the configuration are never applied to WSV. One possible cause is that the event counters are detecting more events than the actual number of blocks committed, and exiting prematurely

[appetrosyan]

The formatting "rose#wonderland@@wonderland" seems odd to me.

[s8sato]

I think so too. Created #2310

[SamHSmith]

I believe this is actually a consensus issue. It is probably the case that sumeragi currently can't handle high load due to some bug. This bug is however only triggered during the benchmark since it is the only time sumeragi is put under high tx load.

[appetrosyan]

Related #2078

[SamHSmith]

This issue no longer happens and so it can be closed.