Issues

• zu2: re-adding an edge that is already there costs a live log record, so a MERGE shaped workload grows the log without bound

  bugstorage
  tamnd/zu#784 · Filter by author tamnd opened on Aug 25, 2026
  1

• zu2: both seqlock readers can return a torn read, because the validating load has nothing under it

  bugperf
  tamnd/zu#782 · Filter by author tamnd opened on Aug 25, 2026
  1

• Closing a connection does not give back everything it took, after the connection has been interrupted or shared

  tamnd/zu#778 · Filter by author tamnd opened on Aug 25, 2026

• zu2: memory_pages bounds nothing while an async write is going, and the overshoot grows with the load

  bugperf
  tamnd/zu#775 · Filter by author tamnd opened on Aug 25, 2026
  2

• Decimal::parse does not check the result against MAX_DIGITS, so a 39 digit decimal parses

  tamnd/zu#774 · Filter by author tamnd opened on Aug 25, 2026

• Decimal is not re-exported from the zu crate, so every client depends on zu-common directly

  tamnd/zu#773 · Filter by author tamnd opened on Aug 25, 2026

• zu2: a scan plane node spends 8 bytes on a header that fits in 4, and pays for it twice

  perfstorage
  tamnd/zu#772 · Filter by author tamnd opened on Aug 25, 2026
  1

• zu2: map_settled makes one VMA per log page, so the mapping count grows with the database

  bugperfstorage
  tamnd/zu#768 · Filter by author tamnd opened on Aug 25, 2026
  3

• zu2: what is left of the anonymous budget once the log pages are mapped

  perfstorage
  tamnd/zu#767 · Filter by author tamnd opened on Aug 25, 2026
  3

• The corpus DECIMAL parser #761 added cannot be reached, and the README says BYTES is still reserved

  tamnd/zu#764 · Filter by author tamnd opened on Aug 25, 2026

• Client test suites name things after reserved words, and only find out when the pin moves

  tamnd/zu#762 · Filter by author tamnd opened on Aug 25, 2026

• memory_pages evicts mapped pages, which gives back no anonymous memory and costs a pread

  perfstorage
  tamnd/zu#759 · Filter by author tamnd opened on Aug 25, 2026
  2

• Every resident page is anonymous memory, so zu2 holds 65x more of it than lmdb and the kernel cannot reclaim any of it

  bugperfstorage
  tamnd/zu#757 · Filter by author tamnd opened on Aug 25, 2026
  3

• go-ycsb: the scan path builds a map a row and throws it away, and it caps workload E at 2.94x

  benchgoperf
  tamnd/zu#750 · Filter by author tamnd opened on Aug 25, 2026
  3

• log::load reads the header and then reads it again, which the cold tier stopped doing in #557

  bug
  tamnd/zu#745 · Filter by author tamnd opened on Aug 24, 2026
  1

• Two thirds of a YCSB scan is spent above the C API, not in the engine

  benchperf
  tamnd/zu#743 · Filter by author tamnd opened on Aug 24, 2026
  3

• zu2_read copies the value the way zu2_scan used to, on the workload C hot path

  bugcapiperf
  tamnd/zu#742 · Filter by author tamnd opened on Aug 24, 2026
  2

• The go-ycsb harness reports 34.7 GiB peak RSS for sqlite at 32 threads and I do not believe it

  bug
  tamnd/zu#741 · Filter by author tamnd opened on Aug 24, 2026

• zu2: a borrowing scan for the C API, since the copy is the largest single cost left on workload e

  enhancement
  tamnd/zu#738 · Filter by author tamnd opened on Aug 24, 2026
  3

• zu2: the harness gets 40 to 47 per cent of the engine's scan throughput, and the rest is unattributed

  bug
  tamnd/zu#737 · Filter by author tamnd opened on Aug 24, 2026
  1

• A bulk load puts a third of the records in the cold tier, and every scan of them costs a pread

  bugperfstorage
  tamnd/zu#736 · Filter by author tamnd opened on Aug 24, 2026
  3

• disk_bytes leaves out the checkpoint file, and answers before compaction has finished

  benchbugstorage
  tamnd/zu#733 · Filter by author tamnd opened on Aug 24, 2026
  1

• zu2 scan throughput stops scaling at about eight threads while lmdb keeps going

  bug
  tamnd/zu#732 · Filter by author tamnd opened on Aug 24, 2026
  3

• A byte string can be stored and cannot be read back, so a column of them is write only

  tamnd/zu#728 · Filter by author tamnd opened on Aug 24, 2026

• the ycsb harness compares adapters as much as engines: zu2's reuses what every rival's allocates, and lmdb's returns memory it does not own

  benchbugperf
  tamnd/zu#726 · Filter by author tamnd opened on Aug 24, 2026
  4