Conflicts and wsrep_sync_wait

[jorang]

Does the use of wsrep_sync_wait reduces the probability of conflicts?

For example, if we have two concurrent inserts on the same table from different nodes, if wsrep_sync_wait is set to 4 or 6, does this avoid the insert conflict, in particular, a conflict that results from a gap lock on a primary key?

Thank you.

[krunalbauskar]

Not necessarily but can.
wsrep_sync_wait is designed to reflect the latest data. So imagine it like a hook that is added when you execute it. But the point is not directly controlled by end-user.

Said that, it is not recommended option to control conflict. It is surely recommended option to fetch refreshed data.

[jorang]

Thank you for your reply. So, is there an effective way of avoid conflicts resulting from a gap lock, besides issuing write statements to only one master?

[krunalbauskar]

You can split your workload. Each node can dealt with different table there-by reducing conflict.
If this is not feasible then I don’t foresee any other way out. It is like even on single server if all user threads are trying to access same table then they need to wait.