Extra RTS log messages

XMLWordPrintableJSON

    • Type: Improvement
    • Resolution: Fixed
    • Priority: Major - P3
    • WT11.2.0, 7.1.0-rc0
    • Affects Version/s: None
    • Component/s: None
    • None
    • 2023-04-04 Bibbidi-Bobbidi-Boo, 2023-04-18 Leviosa Not Leviosa
    • 5

      We can improve logging where wt_rts_btree_walk_btree_apply gets these values, and what the values are:

      • newest_start_durable_ts
      • newest_stop_durable_ts
      • rollback_txnid
      • addr_size
      • write_gen

      There’s a gap where wt_rts_btree_walk_btree_apply decides to skip a tree because either the checkpoint address length is zero or it has timestamped updates and the stable timestamp is 0, and later on when we check if it the btree is modified, has prepared updates, etc. If the latter check fails, we log "tree skipped with durable timestamp" but not any of the other conditions.

      wt_rts_btree_walk_btree_apply again: consider logging whether we decide to truncate history store entries, towards the end of the function.

      rts_btree_walk_page_skip: there are a couple of return (0) paths where we can exit without logging anything, perhaps some DEBUG_3 (or higher) messages are appropriate along these paths.

      wt_rts_btree_abort_updates: consider logging the page type.

      rts_btree_abort_col_fix: the call to rts_btree_abort_insert_list should probably have some logging around whether we decide to call it and why.

      rts_btree_abort_insert_list: This could also use some logging.

      rts_btree_abort_col_fix: This has zero logging right now, and so does the "worker" function rts_btree_abort_col_fix_one. rts_btree_abort_ondisk_kv has good logging, but we’ve lost some info by then, especially if we decide not to call it. The same thing applies to the VLCS and row store versions of that function.

      We have no logging for either the callers of wt_rts_history_delete_hs nor in the function itself. Same for wt_rts_history_btree_hs_truncate.

      wt_rts_history_final_pass isn’t too bad, but it would be nice to have logging at the callsite (this assumes we fix logging in wt_rts_history_btree_hs_truncate as well).

            Assignee:
            Clarisse Cheah
            Reporter:
            Will Korteland
            Votes:
            1 Vote for this issue
            Watchers:
            6 Start watching this issue

              Created:
              Updated:
              Resolved: