Wow!
Performance report (total). Blocks: 101240373. Elapsed time: 42386318 ms (real), 51344322 ms (cpu). Memory usage: 12638244 (current), 13218128 (peak) kilobytes. Free space in SHM: 4181002 kilobytes.
42386 seconds is 11 hours and a bit short of 47 minutes. For a replay from scratch to over 101M, no RAM disk, 8GB SHM and block log on an external USB-C HDD.
Looks like number of blocks grows but the time to replay shrinks :o)
I have an interesting one, resync from scratch with checkpoint at 101M
exact same hardware, same software, same datacenter, same rack, same switch
and even sync started at the exact same time
68315 2025-11-20T10:39:31.750 database.cpp:5937 apply_hardfork ] HARDFORK 28 at block 101319928vs
65696 2025-11-20T09:55:52.858 database.cpp:5937 apply_hardfork ] HARDFORK 28 at block 101319928Full resync took 18-19 hours more or less.. but the thing is that the difference is so big:
43m39s (4%)
Two facotors that comes to my mind: difference between wear of disks (those are far from being bran new, so it might be significant at this point), and the other - difference in "luck" when it comes to finding optimal peers.
Update to myself: I just ran replay on those two machines and it was:
58553vs
58266So still one is consistently slower, but by a very small amount (just 4m47s = 0.5%)