We are running our BOE-XI (R2) clustered platform using a SAN for the Input and Output FRS paths using a UNC.
In our Filestore folders on the SAN there is approx. 104,736 files & 921,786 folders.
The file-count seems correct (all Objects and saved Instances) - but the number of Folders seems very high.
There are almost nine (9) folders for each Objects or saved Instance.
Is this correct? Is is the folder-nesting really that deep - or is it a BUG where the BOE application removes Expired files (Objects/Instances) it leaves the old EMPTY folders behind and does not clean them up.
I am not sure that 9 to 1 is the “Correct” ratio of folders to files but BO does create many sub-folders for each file. Once every 6 months we add the line parameter “-Prune” to the CMS server in CCM then start the service. When CPU for CMS service goes back to 0 then its done - remove parameter and restart CMS. It takes anywhere from 10 to 20 minutes to run. Obviously, the BO system must be unavailable to users.
I believe the large number of directories is a side-effect of their design to come up with unique directory areas to store the documents. Maybe they found that this performs better?
In the back of one of the admin guides, I thought there was an appendix containing line commands for the services. I would like to see BO create a specific Configuration Guide with detail steps on setting up the various parts of BO (SSO, Servers, LDAP,…etc.). A more hands-on guide with how and why to apply settings like -prune.
Ran the clean-up on our TEST environment last night. It took about 2 hours to clean-up 400,000+ empty folder paths.
PROD is probably around 600,000 empty folder paths - but I need to submit an RFC to do anything on PROD.
Having to STOP the FRS to clean it up using “-Prune” is going to be a major issue for some clients. I wonder what the SAP-BOBJ strategy is for cleaning up these empty-FRS folders in high-volume / high-availability environments.
It would be nice if the CMS or the FRS would do this empty folder paths clean-up when there was idle time on the system (or when the object/instance file was DELETED).
(BOB member since 2007-03-23)
(BOB member since 2005-10-02)