SELECT FOR UPDATE/SELECT LOCK IN SHARE MODE does not work in PXC, but are allowed for pxc_strict_mode=ENFORCING

[Grigorijs]

Issuing SELECT FOR UPDATE/SELECT LOCK IN SHARE MODE is working as expected only when used inside one cluster node. If parallel transaction is tried on another node then no limitations are applied on it.

Think right handling would be to deny these operations with ENFORCING pxc_strict_mode, or implement correct handling (so SELECT FOR UPDATE on one node would prevent for example SELECT LOCK IN SHARE MODE on another node)

Current software version is:
ii percona-xtradb-cluster-server-5.7 5.7.14-26.17-1.jessie amd64

[Kenn Takara (kenn-takara)]

Galera does not provide for this kind of cross-cluster locking. With that said, this functionality has been left in to allow the code to work in single-node writer scenarios.

[Grigorijs (grigk)]

How I see it, handling of these requests should be same as for other locking operations (LOCK TABLES, GET_LOCK) - depending on pxc_strict_mode silently allowed, warning issued or denied. As by logic it seems to me that underlying LOCK system is the same for all these requests

[Krunal Bauskar (krunal-bauskar)]

Problem with Select for Update is not always true. So it is better treated as limitation than bug.

[Shahriyar Rzayev (rzayev-sehriyar)]

Percona now uses JIRA for bug reports so this bug report is migrated to: https://jira.percona.com/browse/PXC-1942