Just the other day, one of our developers was having problems getting our California AWS instance back up and running. The day before, we started moving our DW flat files to the instance using the SFTP plug-in. Needless to say, the stock AWS 120 GB drive configuration wasn’t large enough to accommodation all the data files and the HANA database. What made this “strange” is that we have another instance in Oregon with the same flat files and database.
What’s the difference you ask – simple – with our California instance, the tables are row based. In Oregon, the tables are column based! A quick check of the tables showed that we were getting a 12.5 times compression with the column tables – that makes a big difference. We were trying to evaluate query performance of row vs column based tables in similar environments – that will be a topic for another blog.
Unfortunately, our shop is mainly a Windows Server shop and we’re not used to the Linux command line tools. We were also in a state where the HANA instance would not load and HANA Studio could only run in diagnostics mode – that is – you can’t see the alerts.
After some digging around – we learned about the df -lh command. Sure enough the /dev/md0 device with houses the /sap home directory was 100% full. After deleting our flat files – we got back down to 74% usage, so after stopping and starting the HANA instance using Linux – see http://scn.sap.com/community/developer-center/hana/blog/2012/11/26/stop-and-start-rules-for-hana-on-aws everything started up great.
In HANA Studio, we brought up the Administration page and sure enough in the Alerts page, there was the HIGH priority disk full alert. After following the steps in the Alert Details dialog and then going back to the Over view page and clicking on the Disk Full Events link and clearing the events, our instance is in a happy state.
We then went back to the Administration guide http://help.sap.com/hana/hana_admin_en.pdf and learned how to setup alerts – starting at page 70.
1. ROW tables don’t have any compression and COLUMN tables do – 12.5x with one of our sample data sets.
2. Alerts do matter – pay attention to them. Next trick is to figure out how to configure HANA and AWS to send emails.
3. Learn the basic Linux commands – that will be another blog post too 🙂
If you like this post – please rate it 🙂
You can follow me on twitter at http://twitter.com/billramo
Data Architect, Advaiya Inc.