Problem Note 65844: STRESS task fails with "Fatal error in PMPI_Bcast: Other MPI error, error stack: PMPI_Bcast(1478)"
In SAS® High-Performance Risk, a STRESS task might fail with a message like the following in the SAS log while the compute server is sending the ScenarioCF/Value data to the HPRisk Engine:
MPIDI_CH3i_Progress_wait(213).............: an error occurred while handling an event returned by MPIDU_Sock_Wait()
MPIDI_CH3I_Progress_handle_sock_event(437):
MPIDI_CH3_PktHandler_EagerSend(606).......: Failed to allocate memory for an unexpected message.
261895 unexpected messages queued.
Error messages like the following might subsequently occur in the SAS log:
ERROR: The application received an invalid market buffer size for the basecase from the HPRisk Engine.
ERROR: Information could not be read from the HPRisk Engine. The HPRISK procedure must terminate the task.
or
ERROR: Journals were not received from the HPRisk Engine after sending scenario cash flow data to the HPRisk Engine.
ERROR: Information could not be read from the HPRisk Engine. The HPRISK procedure must terminate the task.
This problem occurs because the worker nodes cannot process quickly enough the data that is being sent to them.
Click the Hot Fix tab in this note to access the hot fix for this issue.
With the hot fix applied, you can specify the option CFOBS_IN_BATCH= in the PROC HPRISK statement to regulate the rate at which the compute server sends ScenarioCF/Value data to the HPRisk Engine. For example, you can test by specifying values such as 1000, 10000, 100000, and 1000000 in order to determine an appropriate value of CFOBS_IN_BATCH= option for a particular set of ScenarioCF/Value data.
Operating System and Release Information
SAS System | SAS High-Performance Risk | Solaris for x64 | 3.8 | 4.3 | 9.4 TS1M4 | 9.4 TS1M7 |
Linux for x64 | 3.8 | 4.3 | 9.4 TS1M4 | 9.4 TS1M7 |
64-bit Enabled Solaris | 3.8 | 4.3 | 9.4 TS1M4 | 9.4 TS1M7 |
64-bit Enabled AIX | 3.8 | 4.3 | 9.4 TS1M4 | 9.4 TS1M7 |
Microsoft® Windows® for x64 | 3.8 | 4.3 | 9.4 TS1M4 | 9.4 TS1M7 |
*
For software releases that are not yet generally available, the Fixed
Release is the software release in which the problem is planned to be
fixed.
Type: | Problem Note |
Priority: | high |
Date Modified: | 2021-07-01 08:07:45 |
Date Created: | 2020-04-14 06:49:38 |