Percona Server moved to https://jira.percona.com/projects/PS

  1. Bug #1076215
  2. Comment #1

Comment 1 for bug 1076215

Revision history for this message
Raghavendra D Prabhu (raghavendra-prabhu) wrote : Re: some confusing code in buf_LRU_mark_space_was_deleted

@yinfeng,

I noticed that both the lazy / non-lazy implementations call btr_search_drop_page_hash_index

 if (srv_lazy_drop_table) {
  buf_LRU_mark_space_was_deleted(id);
 } else {
 buf_LRU_flush_or_remove_pages(
  id, evict_all
  ? BUF_REMOVE_ALL_NO_WRITE
  : BUF_REMOVE_FLUSH_NO_WRITE);

 }

buf_LRU_flush_or_remove_pages also calls it.

Regarding btr_search_drop_page_hash_when_freed from fseg_free_extent I noticed that only dict_truncate_index_tree makes call to it in row_truncate_table_for_mysql, though that is done after fil_discard_tablespace

Can you tell us the call-path where you found duplicate calls to btr_search_drop_page_hash_when_freed / btr_search_drop_page_hash_index (gdb bt with drop table should do)?