It seems when using good storage that this thread act as bottleneck in a way that other threads are waiting completion from it a the cpu level when performing initial full node synchronisation.
As a result, Bitcoin is mostly single threaded in that case (the -msghand thread take 100% of a single core leaving other cores idle even on a core duo cpu).
This what currently prevents syncing a full node with -txindex in less than 1 hour.
I have the same problem on my four core machine. the msghand thread is running at 100% and not much else happens.
Closing for now given this is a general ask and not a bug.
@fanquake no, this is a real performance issue as the database grow. Please remove this bottleneck and don’t turn bitcoin core into OpenEthereum.
@ytrezq more than 100 issues have been closed in last 2 weeks is:issue is:closed sort:updated-desc which includes some that were not resolved although I fail to see what would this improve.
You could comment in #25823 or everyone could consider issues have 2 tabs: open (less important) and closed but not resolved (more important) and keep a track of all the issues.
@1440000bytes because msghand is single threaded, parallelism is limited in a way more cores don t means faster.
But yes, if this should be closed, this should be done as uncompleted.