database/openGauss-server
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

endless loop in readrecord when failover

Browse Source
This commit is contained in:
shirley_zhengx
2024-03-06 20:17:05 +08:00
parent cdae0a6ce5
commit 027c925257
1 changed files with 10 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
src/gausskernel/storage/access/transam/xlog.cpp
View File

@ -10282,9 +10282,16 @@ void StartupXLOG(void)
// Allow read-only connections immediately if we're consistent already.
CheckRecoveryConsistency();
ResetXLogStatics();
// Find the first record that logically follows the checkpoint --- it
// might physically precede it, though.
if (XLByteLT(checkPoint.redo, RecPtr) || SS_STANDBY_PROMOTING) {
/* Find the first record that logically follows the checkpoint --- it
* might physically precede it, though.
* When dms is enabled, due to failover and switchover of intercluster could happend frequently,
* checkPoint.redo is equel to RecPtr and checkPoint.redo is last record, so here record which
* next read should be Startptr of checkPoint.redo, instead of Endptr of checkPoint.redo, which
* could be NULL. Otherwise, the process will end in a loop of this function(ReadRecord).
*/
if (XLByteLT(checkPoint.redo, RecPtr) || SS_STANDBY_PROMOTING ||
(XLByteLE(checkPoint.redo, RecPtr) && SS_ONDEMAND_REALTIME_BUILD_NORMAL)) {
/* back up to find the record */
record = ReadRecord(xlogreader, checkPoint.redo, PANIC, false);
} else {