MPX_WORK_UNIT_TIMEOUT Option

Time, in seconds, before a multiplex DQP leader reassigns incomplete distributed work to another DQP worker node.

Allowed Values

0 to 3600 seconds.

DQP work units are typically sized to span only a few seconds. If a worker node goes offline or experiences an unusually high workload, DQP work previously assigned to that worker node is reassigned to another node after the given timeout.

Default

60 seconds

Typically you do not need to change this option from its default value. However, increase this option in rare cases where a query has very large intermediate results that cause individual work units to time out.

Decrease this option if unreliable networks or servers cause distributed work to be lost and the timeout interval is unacceptably long. Note that setting this option too low can cause unnecessary early timeouts.

Scope

Can be set temporary for an individual connection or persistently across server restarts. Requires DBA permissions to set the option. Setting takes effect immediately.