main mainnet:
- in main during initial sync i see:
tail -10000f /erigon-data/mainnet_stepsize4_regen/logs/erigon.log | grep -i collate | grep -v merge
[DBUG] [05-21|04:55:05.340] [agg] collate-build step=9001 collate_workers=2 compress_workers=2 collate_in=5.426286509s build_in=3.964590923s
[DBUG] [05-21|04:57:18.456] [agg] collate-build step=9002 collate_workers=2 compress_workers=2 collate_in=5.610692368s build_in=3.573841385s
[DBUG] [05-21|05:10:22.178] [agg] collate-build step=9003 collate_workers=2 compress_workers=2 collate_in=7.097553171s build_in=3.687605135s
[DBUG] [05-21|05:10:34.477] [agg] collate-build step=9004 collate_workers=2 compress_workers=2 collate_in=7.206177623s build_in=5.092747962s
[DBUG] [05-21|05:10:47.122] [agg] collate-build step=9005 collate_workers=2 compress_workers=2 collate_in=6.48712595s build_in=4.776940959s
[DBUG] [05-21|05:10:57.088] [agg] collate-build step=9006 collate_workers=2 compress_workers=2 collate_in=5.912039343s build_in=4.053565178s
[DBUG] [05-21|05:11:06.896] [agg] collate-build step=9007 collate_workers=2 compress_workers=2 collate_in=5.899953193s build_in=3.907728765s
between 9003 and 9004 - many-many minutes. during that time stepsInDB was > 4
- Also i saw very big
BlockTransactions table (10GB). Which slowly goes down as execution happening (i guess Caplin inserting all batches instead of insertbatch+execbatch).
i guess it's because in main we have statefull agg.SetMaxCollationTxNum() field - which we don't need. i guess it was temporary workaround for "state ahead blocks" issue
Also in logs i see:
[INFO] [05-21|06:11:18.539] [4/6 Execution] serial done in=7.719933007s buf=116.2MB/512.0MB blk=25119199 blks=100 blk/s=12 txs=60.18k tx/s=7.79k gas/s=400.12M stepsInDB=1.64 step=9019.8 alloc=11.2GB sys=23.7GB isForkValidation=false isApplyingBlocks=true
[INFO] [05-21|06:11:18.758] BuildFilesInBackground step=9018 lastInDB=9019 execStep=9019
[INFO] [05-21|06:11:18.758] [snapshots] holding state collation at block snapshot boundary blockSnapshotsStepCompleted=9017 lastCollatableStepInDb=9034
[WARN] [05-21|06:11:20.624] [dbg] SetMaxCollationTxNum step=9035 stack="[aggregator.go:1582 exec3.go:144 stage_execute.go:395 default_stages.go:222 sync.go:500 sync.go:419 executor.go:149 forkchoice.go:507 forkchoice.go:127]"
[INFO] [05-21|06:11:20.624] [4/6 Execution] serial starting from=25119200 to=25119299 initialTxNum=3523327751 initialBlockTxOffset=0 lastFrozenStep=9017 initialCycle=true isForkValidation=false
blockSnapshotsStepCompleted=9017 lastCollatableStepInDb=9034 - somehow big difference. while stepsInDB is small
mainmainnet:between 9003 and 9004 - many-many minutes. during that time stepsInDB was > 4
BlockTransactionstable (10GB). Which slowly goes down as execution happening (i guess Caplin inserting all batches instead of insertbatch+execbatch).i guess it's because in main we have statefull agg.SetMaxCollationTxNum() field - which we don't need. i guess it was temporary workaround for "state ahead blocks" issue
Also in logs i see:
blockSnapshotsStepCompleted=9017 lastCollatableStepInDb=9034- somehow big difference. while stepsInDB is small