What determines the speed of the reindexing-chainstate in the different times? I spend a few days (2 week) for reindexing locally on my new HDD, and I noticed one thing: the speed of reindexing is not constant and sometimes it takes 1 second to execute 10 blocks, blocks flyes through your screen, your CPU is 100% loaded , but sometimes it takes 30 seconds for one block and your CPU take 10-20%. It can happen the same way in earlier or later blocks, it doesn't matter. I know reindexing to exeсute locally and this is not the fault of a weak Internet connection. The blocks are in most cases the same size. Why the reindexing speed is not constant?
1 Answer
Updates to the UTXO are write cached, you will see pauses when this is flushed to disk periodically.
-
Thanks, but in most cases, reindexing is slow and only occasionally accelerates to good speeds and lasts for hours. This is not pauses. Commented Aug 28, 2021 at 22:54
-
1Due to the design of the UTXO cache in Bitcoin Core, you generally get significantly reduced performance just after flushing, as it involves wiping the cache. Depending on the size of your dbcache and the speed of your disk, this effect can be pretty extreme. Commented Aug 29, 2021 at 17:46