ValueError: max() arg is an empty sequence if there is only 1 filled block

[drandreaskrueger]

when very few transactions, and all fit into 1 block:

Send 200 transactions 
...
block 3 | new #TX 200 / 4000 ms =  50.0 TPS_current | total: #TX  201 /  4.6 s =  43.5 TPS_average (peak  is  43.5 TPS_average)
block 4 | new #TX   0 / 4000 ms =   0.0 TPS_current | total: #TX  201 /  8.0 s =  25.1 TPS_average (peak  is  25.1 TPS_average)
block 5 | new #TX   0 / 4000 ms =   0.0 TPS_current | total: #TX  201 / 11.9 s =  16.8 TPS_average (peak  is  16.8 TPS_average)
block 6 | new #TX   0 / 4000 ms =   0.0 TPS_current | total: #TX  201 / 16.2 s =  12.4 TPS_average (peak  is  12.4 TPS_average)
block 7 | new #TX   0 / 4000 ms =   0.0 TPS_current | total: #TX  201 / 19.8 s =  10.1 TPS_average (peak was  12.4 TPS_average)
block 8 | new #TX   0 / 4000 ms =   0.0 TPS_current | total: #TX  201 / 24.1 s =   8.3 TPS_average (peak was  12.4 TPS_average)
block 9 | new #TX   0 / 4000 ms =   0.0 TPS_current | total: #TX  201 / 28.0 s =   7.2 TPS_average (peak was  12.4 TPS_average)
block 10 | new #TX   0 / 4000 ms =   0.0 TPS_current | total: #TX  201 / 32.0 s =   6.3 TPS_average (peak was  12.4 TPS_average)
...
=============================
= blocksDB_diagramming.py
=============================
...
from block 3 to block 3, with 10 empty blocks afterwards

then there is a problem:

./blocksDB_diagramming.py:422: RuntimeWarning: invalid value encountered in double_scalars
  tps=(txs/duration)

Traceback (most recent call last):
...
    blMin, blMax = min(dfs["blocknumber"])+1, max(dfs["blocknumber"][:lastFilledBlock_index])
ValueError: max() arg is an empty sequence

details:

peak TPS single block:
    blocknumber  TPS_1blk  TPS_3blks  TPS_5blks  TPS_10blks  txcount   size  gasUsed  gasLimit   timestamp  blocktime
2             5       0.0        NaN        NaN         NaN        0    586        0  40000000  1551190348        4.0
3             6       0.0        NaN        NaN         NaN        0    586        0  40000000  1551190352        4.0
4             7       0.0        0.0        NaN         NaN        0    586        0  40000000  1551190356        4.0
5             8       0.0        0.0        NaN         NaN        0    586        0  40000000  1551190360        4.0
6             9       0.0        0.0        0.0         NaN        0    586        0  40000000  1551190364        4.0
7            10       0.0        0.0        0.0         NaN        0    586        0  40000000  1551190368        4.0
8            11       0.0        0.0        0.0         NaN        0    586        0  40000000  1551190372        4.0
9            12       0.0        0.0        0.0         NaN        0    586        0  40000000  1551190376        4.0
10           13       0.0        0.0        0.0         NaN        0    586        0  40000000  1551190380        4.0
0             3       NaN        NaN        NaN         NaN      200  27462  5354823  40000000  1551190340        NaN

peak TPS over ten blocks:
   blocknumber  TPS_1blk  TPS_3blks  TPS_5blks  TPS_10blks  txcount   size  gasUsed  gasLimit   timestamp  blocktime
0            3       NaN        NaN        NaN         NaN      200  27462  5354823  40000000  1551190340        NaN
1            4       NaN        NaN        NaN         NaN        0    586        0  40000000  1551190344        NaN
2            5       0.0        NaN        NaN         NaN        0    586        0  40000000  1551190348        4.0
3            6       0.0        NaN        NaN         NaN        0    586        0  40000000  1551190352        4.0
4            7       0.0        0.0        NaN         NaN        0    586        0  40000000  1551190356        4.0
5            8       0.0        0.0        NaN         NaN        0    586        0  40000000  1551190360        4.0
6            9       0.0        0.0        0.0         NaN        0    586        0  40000000  1551190364        4.0
7           10       0.0        0.0        0.0         NaN        0    586        0  40000000  1551190368        4.0
8           11       0.0        0.0        0.0         NaN        0    586        0  40000000  1551190372        4.0
9           12       0.0        0.0        0.0         NaN        0    586        0  40000000  1551190376        4.0

Single block, vs averaged over 10 blocks:
peak( TPS_1blk) = 0.00 
peak(TPS_10blk) = nan

./blocksDB_diagramming.py:422: RuntimeWarning: invalid value encountered in double_scalars
  tps=(txs/duration)
second to last experiment block, averaging:
blocks 3-3, timestamps 1551190340-1551190340, duration 0 seconds, txcount 0, tps nan

Traceback (most recent call last):
  File "./blocksDB_diagramming.py", line 713, in <module>
    load_prepare_plot_save(*params)
  File "./blocksDB_diagramming.py", line 633, in load_prepare_plot_save
    emptyBlocks=EMPTY_BLOCKS)
  File "./blocksDB_diagramming.py", line 562, in diagrams
    tpsAv = tps_plotter(axes[0,0], dfs, blockFrom, blockTo, emptyBlocks)
  File "./blocksDB_diagramming.py", line 488, in tps_plotter
    avgLine(ax, dfs, emptyBlocks, avg, avgTxt)
  File "./blocksDB_diagramming.py", line 451, in avgLine
    blMin, blMax = min(dfs["blocknumber"])+1, max(dfs["blocknumber"][:lastFilledBlock_index])
ValueError: max() arg is an empty sequence

"./blocksDB_diagramming.py $DBFILE $INFOWORD ../$INFOFILE" command filed with exit code 0.

[zhuang-weiming]

I have the same problem on ethermint ,
I searched this problem maybe related with here, the same content here.

I wondering how chainhammer project fininshed stress test about the tx.nonce squence prolem?

[drandreaskrueger]

Do you see the same error message ValueError: max() arg is an empty sequence, or what makes you think this is the same problem?

Here in my case it is probably pretty trivial, I just left this issue as a TODO for myself, so I won't forget about it. Thanks for the reminder, I am fixing it today then.

First step - how to reproduce:

CH_TXS=1 CH_THREADING="sequential" ./run.sh just-testing geth-clique

source ../env/bin/activate
./blocksDB_diagramming.py temp.db Geth ../hammer/last-experiment.json

looking into it now ...

[drandreaskrueger]

Obviously, in a benchmarking situation it is an extreme edgecase to fill only 1 block. But then, when trying to plot the diagrams, calculating averages, etc. was causing three problems, see this commit: 28c9cb4

My solution in the end was simply forbidding the whole case, i.e. not generating any diagram if there are less than 2 filled blocks, see here:

chainhammer/reader/blocksDB_diagramming.py

Lines 631 to 640 in 3c8c3ec

	if TO_BLOCK-FROM_BLOCK>0:
	# fn = diagrams_oldversion(df, FROM_BLOCK, TO_BLOCK, NAME_PREFIX, gas_logy=True, bt_logy=True, imgpath=imgpath)
	fig, axes, dfs, txs, tpsAv = diagrams(NAME_PREFIX, df, FROM_BLOCK, TO_BLOCK,
	emptyBlocks=EMPTY_BLOCKS)
	fn = savePlot(fig, NAME_PREFIX, FROM_BLOCK, TO_BLOCK, imgpath, INFOFILE)
	print ("\ndiagrams saved to: ", fn)
	else:
	print ("\nLess than 2 blocks filled. So: NOT generating diagram!\nPremature ending. You probably do not want this.")
	fn="no-diagram-created-because-less-than-2-filled-blocks"
	tpsAv=0

Solved. Closing this. Thanks for the reminder.

Good luck finding a solution for your problem.