Prune undermines the dbcache. #11315

1. 
   gmaxwell commented at 0:20 am on September 13, 2017: contributor

   17:13:01 < esotericnonsense> Aaaaand: in particular what was happening for me was that the utxo cache seemingly was being dropped on every prune event 17:13:18 < esotericnonsense> so with prune off, it’d build up to 2-3GB or more, with it on, it’d never go above 300mb or so

   We should probably add the dbcache size to the amount of extra blockdata that prune lets the blocks and disk take up over the pruning size… so that pruning doesn’t undermine the user’s dbcache settings.

2. fanquake added the label Resource usage on Sep 13, 2017
3. 
   esotericnonsense commented at 0:25 am on September 13, 2017: contributor

   Additional information on how this happened:

   0.15.0rc3 running on a Linode VPS with prune=6480 dbcache=8000.

   debug.log would indicate cache dropping on every prune event as stated.

   When raising prune to 64800, the behaviour stopped happening until hitting the new prune target.

   Sounds like gmaxwell’s identified the issue already though.

4. 
   Sjors commented at 9:52 am on May 23, 2018: member

   I did some benchmarking in #12404 (comment). So far it appears that reducing the number of flush events if helpful, but reducing them too much appears counter productive, for reasons I don’t understand yet.
5. 
   Sjors commented at 4:32 pm on November 19, 2019: member

   Continued discussion in #15265

Contributors
gmaxwell esotericnonsense Sjors

Labels
Resource usage