Issues
- Massive concurrent inserts on a node with super_read_only enabled can bypass read only settingsPXC-4304Resolved issue: PXC-4304
- Crash when running several MTR tests: unallowed state transition: disconnected -> initializedPXC-4195Resolved issue: PXC-4195Anton Matvienko
- Crash after setting wrong gcs.recv_q_soft_limit valuePXC-4185Resolved issue: PXC-4185
- Build from the latest source instructions refer non existent PXB tarballsPXC-4169Resolved issue: PXC-4169patrick.birch
- pxc_strict_mode does not detect version mismatchPXC-4163Resolved issue: PXC-4163Kamil Holubicki
- wsrep_cluster_size at 0 when syncing from earlier versionPXC-4162Resolved issue: PXC-4162Kamil Holubicki
- Setting up multi source replication with a node from percona xtradb cluster as a slavePXC-4153Resolved issue: PXC-4153
- Build fails with WITH_ROCKSDB=ONPXC-4149Resolved issue: PXC-4149Varun Arakere Nagaraju
- After upgrade to 8.0.30 errors: MDL conflict db= table= ticket=10 solved by abortPXC-4148Resolved issue: PXC-4148Zsolt Parragi
- PXC bootstrap service changes incomplete for systemdPXC-4143Resolved issue: PXC-4143Kamil Holubicki
- WSREP applier threads fail to modify read only schemasPXC-4137Resolved issue: PXC-4137Venkatesh Prasad
- wsrep_desync is not set correctly after FTWRLPXC-4135
- [MTR] The test galera_3nodes_nbo.galera_nbo_master_non_prim_failure sporadically fails on JenkinsPXC-4125
- PXC node replica stalls with asynchronous parallel replicationPXC-4116Resolved issue: PXC-4116Zsolt Parragi
- Inconsistent behavior of SET SESSION binlog_format='STATEMENT';PXC-4113
- Document change to TimeoutStartSec for mysql.servicePXC-4110Resolved issue: PXC-4110patrick.birch
- state transfer with SSL disabled in wsrep_provider_options crashes Receiver and Donor nodePXC-3936Resolved issue: PXC-3936Anton Matvienko
- PXC to PXC async replication broken with "Node has dropped from cluster" after pt-table-checksumPXC-2568Resolved issue: PXC-2568
- wsrep_sync_wait documentation is wrong.PXC-2284Resolved issue: PXC-2284alina.derkach
- LP #1250380: Handle full tables more gracefullyPXC-924Resolved issue: PXC-924Anton Matvienko
20 of 20
Massive concurrent inserts on a node with super_read_only enabled can bypass read only settings
Done
General
Escalation
General
Escalation
Description
Environment
None
AFFECTED CS IDs
CS0038757
Details
Assignee
UnassignedUnassignedReporter
Jaime SicamJaime SicamNeeds QA
YesFix versions
Affects versions
Priority
Medium
Details
Details
Assignee
Unassigned
UnassignedReporter
Jaime Sicam
Jaime Sicam
Needs QA
Yes
Fix versions
Affects versions
Priority
MediumSmart Checklist
Smart Checklist
Smart Checklist
Created September 10, 2023 at 4:53 AM
Updated March 6, 2024 at 8:36 PM
Resolved September 12, 2023 at 8:22 AM
Activity
Kamil HolubickiSeptember 12, 2023 at 8:22 AM
I was able to reproduce it locally on 8.0.32.
I confirm that the issue was fixed in 8.0.33. The fix was done by the above-mentioned commit https://github.com/percona/percona-xtradb-cluster/commit/d780c6e01778a0e1037399c7733e68b1c1128524
The issue was that read_only variable was silently/internally modified (set to true) for a short period of time during query execution.
Kamil HolubickiSeptember 11, 2023 at 9:00 AM
AWS instance, Ubuntu 22.04
Checked on versions 8.0.30 and 8.0.32 (from Percona Distribution for PXC)
Setup with 1 node, steps as above.
Not reproducible.
Venkatesh PrasadSeptember 11, 2023 at 7:20 AM
This may have possibly been fixed in 8.0.33 by commit https://github.com/percona/percona-xtradb-cluster/commit/d780c6e01778a0e1037399c7733e68b1c1128524
I cannot repeat this on PXC 8.0.33 but I was able to reproduce this on PXC 8.0.30 and PXC 8.0.32.
On one node, create this table and then enable super_read_only:
Then, on mysqlslap, run inserts concurrently on this table:
At first, you will see these messages that it's not permitted to insert on this node:
However, after a short while the errors stop spawning and the table gets populated:
From here on, you can also perform your own inserts even if super_read_only is still enabled: