pt-table-checksum 2.1.4 miscategorizes Percona XtraDB Cluster-based slaves as cluster nodes
Bug #1063912 reported by
Brian Fraser
This bug affects 1 person
| Affects | Status | Importance | Assigned to | Milestone | |
|---|---|---|---|---|---|
| Percona Toolkit moved to https://jira.percona.com/projects/PT |
Fix Released
|
Critical
|
Brian Fraser | ||
Bug Description
Two bugs:
cluster -> PXC-based slave
The slave is tagged as a cluster node and slave lag isn't checked, even though it was wsrep_on OFF. This is because the code currently just looks for wsrep_on and doesn't check the value.
cluster1 -> cluster2
The second cluster is a slave of the first, but even if the above bug is fixed, we don't check slave lag because we assume that if wsrep_on is ON, then every cluster node will belong to the same cluster. A good 80% solution is checking that wsrep_cluster_name is the same for both nodes, but that'll still fail for clusters with the same name (i.e., the sandboxes in portable-test-suite as of now).
Related branches
lp:~percona-toolkit-dev/percona-toolkit/partial-fix-1063912-ptc-pxc-slaves
- Daniel Nichter: Approve
-
Diff: 905 lines (+586/-33)11 files modifiedbin/pt-query-digest (+208/-6)
bin/pt-stalk (+113/-4)
bin/pt-table-checksum (+30/-7)
lib/Cxn.pm (+3/-1)
lib/RowChecksum.pm (+5/-0)
t/lib/Pingback.t (+67/-7)
t/lib/RowChecksum.t (+45/-4)
t/pt-stalk/plugin.t (+74/-0)
t/pt-stalk/samples/plugin001.sh (+21/-0)
t/pt-table-checksum/bugs.t (+16/-0)
t/pt-table-checksum/run_time.t (+4/-4)
lp:~percona-toolkit-dev/percona-toolkit/fix-1062563-1063912-ptc-pxc-bugs
- Daniel Nichter: Approve
-
Diff: 1190 lines (+897/-95)10 files modifiedbin/pt-config-diff (+0/-15)
bin/pt-kill (+8/-19)
bin/pt-online-schema-change (+8/-19)
bin/pt-table-checksum (+144/-25)
lib/Cxn.pm (+0/-15)
lib/Percona/XtraDB/Cluster.pm (+123/-0)
lib/Sandbox.pm (+83/-2)
t/lib/Percona/XtraDB/Cluster-no-PXC.t (+88/-0)
t/lib/Percona/XtraDB/Cluster.t (+201/-0)
t/pt-table-checksum/pxc.t (+242/-0)
| summary: |
pt-table-checksum 2.1.4 miscategorizes Percona XtraDB Cluster-based - slaves as cluster codes + slaves as cluster nodes |
| Changed in percona-toolkit: | |
| status: | New → Triaged |
| tags: | added: percona-xtradb-cluster pt-table-checksum |
| Changed in percona-toolkit: | |
| importance: | Undecided → Critical |
| milestone: | none → 2.1.5 |
| assignee: | nobody → Brian Fraser (fraserbn) |
| description: | updated |
Revision history for this message
| Daniel Nichter (daniel-nichter) wrote | #1 |
| Changed in percona-toolkit: | |
| status: | Triaged → Fix Committed |
| Changed in percona-toolkit: | |
| status: | Fix Committed → Fix Released |
Revision history for this message
| Shahriyar Rzayev (rzayev-sehriyar) wrote | #2 |
To post a comment you must log in.
Clarification on "cluster -> PXC-based slave". This happens when a node in the cluster has a traditional replication slave attached to it. That slave can be a Percona XtraDB Cluster instance but not being used as a cluster node, so wsrep_on=OFF.