search

ovis-hpc / ldms

OVIS/LDMS High Performance Computing monitoring, analysis, and visualization project.
https://github.com/ovis-hpc/ovis-wiki/wiki
Other
100 stars 52 forks source link

[b4.4] static decomposition: yield a config error for "op" without "group" #1520

Closed narategithub closed 1 week ago

narategithub commented 1 week ago

When "op" was present in the static decomposition configuration but "group" was not defined, the static decomposition skipped row cache creation and continue working with row_cache being NULL. Consequently, when decompose() was later called, ldmsd terminated with SIGSEGV.

This patch add a sanity check in config() so that the static decomposition yields an error when "group" is missing in the presence of "op".

narategithub commented 1 week ago

This PR is for issue #1512 .

narategithub commented 1 week ago

I tested this with the missing "group" in the presence of "op". ldmsd exited due to configuration error as follows:

Thu Nov 14 21:54:37 2024: INFO      : Listening on sock:411 using `sock` transport and `none` authentication
Thu Nov 14 21:54:37 2024: WARNING   : 'interval' is begin deprecated. Please use 'reconnect' with 'prdcr_add' or 'prdcr_start*' in the future.
Thu Nov 14 21:54:37 2024: WARNING   : 'interval' is begin deprecated. Please use 'reconnect' with 'prdcr_add' or 'prdcr_start*' in the future.
Thu Nov 14 21:54:37 2024: ERROR     : store_csv_static_test_sampler: schema "ldms_test_op" requires "group" due to the presence of "op".
Thu Nov 14 21:54:37 2024: ERROR     : msg_no 6: error 22: store_csv_static_test_sampler: schema "ldms_test_op" requires "group" due to the presence of "op".

Thu Nov 14 21:54:37 2024: ERROR     : Configuration error at line 6 (ldmsd.conf)
Thu Nov 14 21:54:37 2024:           : LDMSD_ LDMS Daemon exiting...status 22, Error 22 processing configuration file 'ldmsd.conf'
Thu Nov 14 21:54:37 2024:           : LDMSD_ cleanup end.
narategithub commented 1 week ago

I just found @tom95858 's commit on main that addressed this issue. I'm closing this PR in favor of PR #1521 .