Our DM system running separate Cassandra on every DMA is configured to offload trend data to Central database which is MySQL.
Recently our Central database suffered an outage of 27 hours and even though the data was cached locally on DMs and re-inserted into the Central DB, it got lost anyway due to other reasons.
As the issues were fixed, we’d like to re-import the data for specific elements/parameters in the specific time period from Cassandra to the Snapshot DB. The resolution is one point per 5 minutes, so about 324 data points per parameter (for 27 hours).
A way is to generate the .csv import file and store it in “c:\Skyline DataMiner\System Cache\Offload\Data\” that gets imported fine.
We are able to run a simple query in DevCenter per one parameter, export the data and manually produce the tab-delimited .csv with some data modification, which gets imported successfully. However, it is not feasible to do this for hundreds of elements with ~2-5 parameters per element.
Has anyone encountered similar request before?
Is there a way to automatically re-generate the .csv import files or even direct MySQL INSERT queries for specific Elements, Parameters for a specific Time Period to re-offload the historical trend data to Central MySQL Database?
Or any other ideas how to back-fill the Central DB with trend data?
Thanks for your help.
Hi Martin,
Unfortunately regenerating these .csv offload files isn't currently part of the standard feature set of DataMiner software. As you pointed out we indeed generate offload files at runtime and will validate if these get inserted successfully into the central database.
In the event of a connection failure towards the central database, the DMA's will locally cache the offload files on the filesystem until the central database connection is restored. At that point the system will automatically resume offloading of the offload files.
In case these insert query itself would fail, we will move the offload files to a "failure" subfolder on the local filesystem of the DMA. This is done to prevent this file from blocking further processing of the offload files. If you as admin resolve the insert issue, you can proceed with moving those offload files from the failure folder back towards their original location where they will be automatically processed again by the offload sequence which runs every 5 minutes (by default)
If you would no longer have access to these offload files and you need to restore the missing dataset on the central database, you will need to create an application which can request the trend data locally from the DMS for the missing data period and have this application populating the necessary .csv offload files.
The easiest way forward would then likely be to create an application which can connect to our DataMiner v1 NBI (SOAP, JSON) and use the get trend data methods on this interface to manually compose the offload .csv files again.
This should allow you to recover the missing information from the DMS and insert it still into the central database.