Details
-
Type:
Bug
-
Status: Resolved (View Workflow)
-
Priority:
Critical
-
Resolution: Fixed
-
Affects Version/s: 26.0.0
-
Fix Version/s: Meridian-2016.1.24, Meridian-2017.1.24, Meridian-2018.1.20, Meridian-2019.1.9, 26.1.2, 27.0.0
-
Component/s: Data Output - Newts
-
Security Level: Default (Default Security Scheme)
-
Labels:None
-
Environment:ems15: CentOS 7, long-running 15-minute RRD collections being converted to Newts
-
Sprint:Horizon 2020 - May 13, Horizon 2020 - May 27
Description
Importantly, the system where we're seeing this problem is collecting on a 15-minute interval rather than the default 5-minute one. It's also worth mentioning that we had to do some custom scripting to convert the RRD repository to store-by-foreign-source.
After swinging this system's persisting to Newts, tripling the default heartbeat setting for Newts fetches, and running the RRD-to-Newts converter, we expected to see resource graphs of historical time-series data gradually fill in as the converter ran, eventually meeting up with the ongoing collection. In reality, the historical data never became visible in resource graphs, despite the data from ongoing collection being visible there.
The customer took the initiative to start doing some queries directly against the Cassandra cluster, and found that historical samples were being inserted on 60-minute boundaries rather than 15-minute ones.
