Issues
- Add documentation on Incremental backup with page tracking could fail if it's not able to copy the system tablespacePXB-3325Resolved issue: PXB-3325patrick.birch
- Assertion failure: page0cur.cc:1177:ib::fatal triggeredPXB-2895Resolved issue: PXB-2895Rahul Malik
- Assertion failure: log0recv.cc:2190:!page || page_is_comp(page) == dict_table_is_comp(index->table)PXB-2894Rahul Malik
- Incremental backup crash: Assertion failure: read_filt.cc:236:page_lsn < incremental_lsn || page_lsn >= incremental_start_checkpoint_lsnPXB-2687Satya Bodapati
- PXB gets stuck during incremental backup prepare of encrypted tablesPXB-2681Resolved issue: PXB-2681Rahul Malik
- Incremental backup prepare crash for encrypted tables with page tracking enabledPXB-2680Resolved issue: PXB-2680Rahul Malik
- Incremental prepare crash for create/drop full text index and page tracking enabledPXB-2659Rahul Malik
- Pagetracking return null data if checkpoint LSN is less than page tracking start LSnPS-8710
Add documentation on Incremental backup with page tracking could fail if it's not able to copy the system tablespace
Description
Environment
AFFECTED CS IDs
Details
Assignee
patrick.birchpatrick.birchReporter
Jaime SicamJaime SicamLabels
Needs QA
NoIn progress time
1.1Time tracking
No time logged1d 1h remainingComponents
Sprint
NoneFix versions
Affects versions
Priority
Medium
Details
Details
Assignee
Reporter
Labels
Needs QA
In progress time
Time tracking
Components
Sprint
Fix versions
Affects versions
Priority
MediumSmart Checklist
Smart Checklist
Smart Checklist
Activity
Satya BodapatiSeptember 4, 2024 at 9:49 AMEdited
Unfortunately yes. We have added the current bug to the backlog. We will take care of it in the future to remove the limitation.
Francisco Miguel BieteSeptember 4, 2024 at 8:00 AM
Am I to understand that page tracking is not supported with multi-ibdata tablespace with XtraBackup?
I don’t see that limitation with MySQL Enterprise Backup
Satya BodapatiAugust 13, 2024 at 3:44 PM
Someone asked on the support ticket (may be) , why are the ibdata pages being tracked?? Thats a good question.
It is because of innodb_change_buffering. The change buffer still exists in ibdata*. Undo is in separate tablespaces by default. So another way to workaround this bug would be to disable innodb change buffer by using innodb_change_buffering=none.
Satya BodapatiAugust 13, 2024 at 3:11 PM
Updates: am able to reproduce the bug and record it in rr.
To me it looks this multi-file tablespace (idata*) is causing some confusion wrt to page_number and the actual file offset.
Work in progress..
To reproduce the error consistently, the environment should have:
(1) More than one system tablespace
(2) sysbench running OLTP RW with tables that have secondary indexes
(3) Looping concurrent transactions that delete all rows on the table but then rollback right after
MySQL configuration:
Instructions on populating tables with sysbench and starting it:
Running Xtrabackup full backup followed by successive incremental backups:
At this point incremental backups will run successfully. To get the incremental backup to fail, run the following in different sessions:
Session 1:
Session 2:
Session 3:
Then, incremental backups will fail: