Opened 15 years ago

Closed 15 years ago

Last modified 15 years ago

#911 closed Defect (fixed)

BOINC core hangs when there is active CUDA tasks in queue but CUDA anabled GPU removed

Reported by: Raistmer Owned by: davea
Priority: Blocker Milestone: Undetermined
Component: Client - Daemon Version: 6.6.20
Keywords: Cc:

Description (last modified by Ageless)

Please, see this thread for full diagnostic of this issue.

In short, BOINC core hangs badly when there are active CUDA tasks (anonymous platform used) but no CUDA GPU. It does this after changing CUDA task status from "suspended" to "waiting to run". BOINC manager then restarts core from prev saved state with CUDA task disabled.

Note, thet BOINC correctly determined that there is no CUDA GPUs in system. Moreover, using <no_gpu> tag leads to exactly the same behavior.

There are some crash dumps in logs files.

Attachments (3)

BOINC_hung_after_GPU_replacement_logs.rar (12.1 KB) - added by Raistmer 15 years ago.
BOINC log files, renamed and RARed
BOINC_hung_after_GPU_replacement_logs.zip (14.9 KB) - added by Nicolas 15 years ago.
Reuploaded logs in free format
BOINC_core_hang_bug_additional_logging_enabled.rar (42.0 KB) - added by Raistmer 15 years ago.
Additional logging enabled by cc_config as recommended in related thread

Download all attachments as: .zip

Change History (7)

Changed 15 years ago by Raistmer

BOINC log files, renamed and RARed

comment:1 Changed 15 years ago by Ageless

Description: modified (diff)

Reformatted link.

comment:2 Changed 15 years ago by Ageless

Description: modified (diff)

Added space.

Changed 15 years ago by Nicolas

Reuploaded logs in free format

Changed 15 years ago by Raistmer

Additional logging enabled by cc_config as recommended in related thread

comment:3 Changed 15 years ago by davea

Resolution: fixed
Status: newclosed

(In [18284]) - client: when parsing app_info.xml, make sure coprocessors are present.

fixes #911

comment:4 Changed 15 years ago by romw

(In [18300]) - client: when parsing app_info.xml, make sure coprocessors are present.

fixes #911

client/

cs_statefile.cpp

Note: See TracTickets for help on using tickets.