After all, I have confirmed that this fixes the problem on crash
recovery of hot-standby botfor 9.3 and HEAD and no problem was
found except unreadability :(

By the way, I moderately want to fix an assertion message to a
ordinary one. Details are below.

The server stops with following message during restarting after
crash requesting archive recovery when the WAL has been produced
with the wal_level below WAL_LEVEL_HOT_STANDBY.

| TRAP: FailedAssertion("!(((oldestActiveXID) != ((TransactionId) 0)))", File: 
"xlog.c", Line: 6799)
| LOG:  startup process (PID 7270) was terminated by signal 6: Aborted

Surely this is the consequence of illegal operation but I think
it is also not a issue of assertion - which fires on something
wrong in design or quite rare cases(this case ?). So it might be
better to show message as below on the case.

| FATAL:  Checkpoint doesn't have valid oldest active transaction id
| HINT:  Reading WAL might have been written under insufficient wal_level.

This could do in this way,

diff --git a/src/backend/access/transam/xlog.c 
index e3d5e10..bb6922a 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -6789,7 +6789,13 @@ StartupXLOG(void)
                        if (wasShutdown)
                                oldestActiveXID = 
PrescanPreparedTransactions(&xids, &nxids);
+                       {
                                oldestActiveXID = checkPoint.oldestActiveXid;
+                               if (!TransactionIdIsValid(oldestActiveXID))
+                                       ereport(FATAL,
+                                                       (errmsg("Checkpoint 
doesn't have valid oldest active transaction id"),
+                                                        errhint("Reading WAL 
might have been written under insufficient wal_level.")));
+                       }
                        /* Tell procarray about the range of xids it has to 
deal with */

What do you think about this? Feel free dumping this if you feel
negative on this.


Kyotaro Horiguchi
NTT Open Source Software Center

Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:

Reply via email to