Description of problem:
RegisterVdsQuery is called periodically and updates vds_static, this changes the _update flug inside vds_static and makes ETL getting the information (which was not changed) again and again.
This problem was reported after ETL consumed 100% CPU
Analysis, done by Juan, showed that ETL is periodically changing host configuration, as a result a query based on the host_configuration view
is executed frequently and loading the CPU.
Juan also found that the query is not optimized, a patch for that is already pushed to Gerrit (see : See Also)
It is worth to mention also that this occurred on a system with no RHEV-H , even though , the RegisterVdsQuery that is relevant only for RHEV-H was running.
Version-Release number of selected component (if applicable):
3.0
How reproducible:
Large systems with history module installed and active ETL
Steps to Reproduce:
1.
2.
3.
Actual results:
Expected results:
RegisterVdsQuery should not be called without a reason
The _update flag should be updated only when vds is edited
Additional info:
please note kaul said there are no rhev-h in this environment, so need to revisit the assumption of registerVdsQuery which is supposed to be called only by vdsm_reg (could be old data in the environment as well - easy to check since we have a host type, logs(?), etc.)