ceph/doc/dev/osd_internals/pg.rst

   1 ====
   2 PG
   3 ====
   4
   5 Concepts
   6 --------
   7
   8 *Peering Interval*
   9   See PG::start_peering_interval.
  10   See PG::acting_up_affected
  11   See PG::RecoveryState::Reset
  12
  13   A peering interval is a maximal set of contiguous map epochs in which the
  14   up and acting sets did not change.  PG::RecoveryMachine represents a
  15   transition from one interval to another as passing through
  16   RecoveryState::Reset.  On PG::RecoveryState::AdvMap PG::acting_up_affected can
  17   cause the pg to transition to Reset.
  18
  19
  20 Peering Details and Gotchas
  21 ---------------------------
  22 For an overview of peering, see `Peering <../../peering>`_.
  23
  24   * PG::flushed defaults to false and is set to false in
  25     PG::start_peering_interval.  Upon transitioning to PG::RecoveryState::Started
  26     we send a transaction through the pg op sequencer which, upon complete,
  27     sends a FlushedEvt which sets flushed to true.  The primary cannot go
  28     active until this happens (See PG::RecoveryState::WaitFlushedPeering).
  29     Replicas can go active but cannot serve ops (writes or reads).
  30     This is necessary because we cannot read our ondisk state until unstable
  31     transactions from the previous interval have cleared.