Due to an old and neglected Ubuntu 18.04 server, without support for PHP 7.3, my LibreNMS install has been down since the v1.70.0 release updated the minimum PHP requirement. Rather than upgrade the existing server, I’ve completed a new Ubuntu 20.04.1 install & followed the documentation for a new LibreNMS install and migration.
Immediately following the migration everything appears OK, full graph history etc is still available, until the poller first runs, at which point a large number of my graphs seem to be reset and lose all historic data. It’s mainly disk usage history that I’m interested in.
The data that has been reset doesn’t appear to be consistent against one specific target, or operating system. i.e. I have two file servers, both running Windows Server 2016 Core, on the same virtual hosts:
- A has reset memory data, but B has retained it
- Both have retained their CPU data
- both have reset all disk usage data, for all disks
I did notice the FAQ for migrating mentioned the RRDs potentially needing migrating if you’re changing CPU architecture, however I don’t believe this should apply to me as this is just a new VMware VM on the same host.
I also noticed that the v1.70.0 release post mentions that memory polling has been reworked and may result in loss of history, but I assume that shouldn’t affect storage usage, which is what I’m most interested in retaining history for.
I’ve already cleared the data and re-processed the migration copying the old data across again, and I got either the same or similar results, I didn’t keep specific enough notes to confirm whether the items which were/weren’t reset were identical or not.
Any advice / assistance in troubleshooting this would be appreciated.