Closed belforte closed 5 years ago
DB status became OK after Diego turned off the DBSPublisher for I discussed this with @lecriste @dciangot and @todor-ivanov and we decided to leave current query to ASOLESS since it works and to create a new, ad hoc, one to be used for asoprod1 and to deploy the modified REST only on a dedicated REST instance to be used only for asoprod1 for a week or so until it is drained. @mmascher if you have the time and strength to read, understand, and in case advice... thanks !
this is only being done in ad-hoc branch https://github.com/dmwm/CRABServer/tree/ForAsoprod1
If we ever want to use multiple asoworker names we may have to do something, but not clear what.
this was solved by creating the ad-hoc branch and using it in a server dedicated to remaining ASOv1 work. ASOv1 is now off, forever.
Introducing a second ASO publishers for asoworker=asoless resulted in Oracle not being able to optimize the execution plan fo the query and when asoworker=asoprody query time explode. See discussion upon report from CERN DBA in: https://cern.service-now.com/service-portal/view-incident.do?n=INC1930071&view=itil&width=default
Here's relevant text
When asked more specifically as to what to change, Frank offered the following: