Hello, I'm wondering if there's any supported method for moving Inventory from sqlite to MS SQL?
The long version: After many hours of watching inventory and the sqlite database directory I've begun to notice a problem with it in my environment. Anytime there's a large number of computers being updated (1000+) by something in deploy, Inventory begins to get bogged down. What I mean by that is computers will start to take longer, and longer to complete a scan. Once I have over 300-400 being scanned at once its snowballs into 1000+ and after an hour computers will start to time out and never finish the scan. What else I've noticed, during these large number of scans, is that the Database.db-wal file (in the database directory) is never left alone to write to the database. The db-wal file grows and grows until it's several GB in size (note: my database is only 400ish MB). My understanding is that the db-wal file wants to write to the database after roughly 1000 pages or 4MB but cannot if it's being continually accessed. If I stop the Inventory service, let the db-wal file catchup, and then start the service again computers scan VERY fast (a few seconds). After a little while, when the db-wal file is 2-3 times larger than the database, again, everything has slowed way down. I would like to at least see if moving my Inventory database to MSSQL would solve this problem. I'm also open to any other idea to solve this problem.
A little more information:
- I've tried small and large number of concurrent scans. Large actually seems to be better (500-1000).
- I only have two scans being used currently: one runs every 24 hours, the other runs after a successful deployment.
- I’m using the provided Collection Libraries and Auto Deploy features for most of these large updates, so dividing them into smaller collections to spread the scanning load isn’t any easy solution.
- Deploy and Inventory run on the same box.
- Our VM environment is not limited by resources and the storage is a Nimble 460.
Thank you in advance!
Please sign in to leave a comment.