Closed PaulVMo closed 2 years ago
This is probably correct, but I don't remember why I made this change.
This is probably correct, but I don't remember why I made this change.
IIRC you were experimenting with taking greater advantage of the cache in node
or something else that was also using core
.
This reverts commit 22302c97c48ea27380086dd284de268609ecce62 in attempt to fix a memory leak. While in consensus group, the memory usage for a validator groups and does not return to pre-consensus levels once elected out. I tracked this to large ETS usage after CG (e.g., 3.8 GB on a validator which had a CG run since last reboot which is 10X that of a validator that has not been in CG since reboot). Much of that space is occupied multiple copies of
__blockchain_hex_unclipped_tbl
and__blockchain_hex_clipped_tbl
(one each per epoch in CG).