Re: hot_standby_feedback vs excludeVacuum and snapshots

On Fri, Jun 8, 2018 at 1:53 PM, Simon Riggs <simon@2ndquadrant.com> wrote:
>
> So the attached patch fixes both the bug in the recent commit and the
> one I just found by observation of 49bff5300d527, since they are
> related.
>
> StandbyReleaseOldLocks() can sweep in the same way as
> ExpireOldKnownAssignedTransactionIds().
>

-StandbyReleaseOldLocks(int nxids, TransactionId *xids)
+StandbyReleaseOldLocks(TransactionId oldxid)
 {
  ListCell   *cell,
     *prev,
@@ -741,26 +741,8 @@ StandbyReleaseOldLocks(int nxids, TransactionId *xids)

  if (StandbyTransactionIdIsPrepared(lock->xid))
  remove = false;
- else
- {
- int i;
- bool found = false;
-
- for (i = 0; i < nxids; i++)
- {
- if (lock->xid == xids[i])
- {
- found = true;
- break;
- }
- }
-
- /*
- * If its not a running transaction, remove it.
- */
- if (!found)
- remove = true;
- }
+ else if (TransactionIdPrecedes(lock->xid, oldxid))
+ remove = true;


How will this avoid the bug introduced by your recent commit
(32ac7a11)?  After that commit, we no longer consider vacuum's xid in
RunningTransactions->oldestRunningXid calculation, so it is possible
that we still remove/release its lock on standby earlier than
required.

-- 
With Regards,
Amit Kapila.
EnterpriseDB: http://www.enterprisedb.com