So far I have noticed this on two devices, I’ll keep looking but both devices were at a location that was w/o power and I’m assuming LibreNMS ran some type of discovery during that time and is no longer pulling historical data. However, other devices that were also at that site (diff type of device) did not lose their historical data.
Why does this happen and can it be corrected in an update?
According to an LLM, the data is not there and would need to be recovered from the folder where the data is store prior to aug from a backup.
Are you on the monthly release? based on the language of this post, you probably should be.
You didn’t give enough data to figure out what issue you had and no one else seemingly reported an issue. For example, you could be talking about a specific device os which had a contributor change the sensor index for something (which we try to catch in PRs but sometimes miss).
If you had data before often it is an rrd file with a different name than LibreNMS expects (usually due to a rename which should not happen normally).
I am on the default release schedule.
There was a power outage at this location, that’s when the graph data deleted or a new file was created when power was restored and the device was rediscovered, checked back in, etc.
However, it did not happen for all devices at that site that were ‘not scanned’ during the power outage, which is odd to me.
Thanks.
“default” is daily. You are presented with information at install time to help inform you if you want to change that setting.
Your info is still too generic to give any advice. Which OS(es)? Debug discovery output? rrd directory listing? device event logs? Anything that looks suspicious or helpful?
Default means you don’t change anything on install and just follow the normal instructions. Had I changed it to something else then it would not be default. That’s my baseline for default. I guess I am on daily updates.
Which OS is librenms on? Or which OS are the devices?
I don’t know how to give you discovery output or RRD directory listing. All I know is that I’ve been using LibreNMS for almost 10 years and 2026 has been the worst year for this server. I’ve lost power plenty of times in those 10 years between multiple sites and I have never had RRD graphs lose their data.
I guess I’ll need to live it no longer having historical data as it doesn’t seem like we’ll be gettin anywhere here.
I do appreciate the replies, thanks.
I just found another device that lost all of its data at the end of August 2026. No power outage, just started from scratch on September 1. Uptime of this device is 3 weeks. Device was added 3 years ago. Last discovery was 2 hours ago. It is a network switch with a static IP at a remote location. I have other devices at this same remote location that display their historical data w/o issues.
There appears to be an issue with LibreNMS somewhere and/or with certain devices. My devices that are losing their data don’t share the same brand, model, etc they are all different brands/models/etc.
Interesting.
According to the LLM I would need to recover RRD graphs from a backup. Why is LibreNMS randomly erasing/overwriting RRD graph files? This is becoming an issue more and more.
@dt-lib you are the only one reporting any issue, and you haven’t given me any debug info so I can’t even really guess.
Since this isn’t happening with all devices especially during the same power outage, I would imagine that it is somewhat random. I’m not sure how much debug info I could paste over here since I would need to clean the output and until I see the output I’m not sure how much I’d need to edit.
However, something is happening as this can’t be normal.
Thanks.
Edit- Also, this is recent. I’ve been running librenms in the same environment for 5-7 years now and I’ve encountered many power outages, over that time period, where data was not lost once power was restored and libre started polling the devices again.
I am also having another issue and not sure if it is related to the power outage scenario.
I have cellular routers in the field and sometimes they lose connection based on their location (the devices are not always in the same spot). Some of my devices that I have graphs configured for are showing 'no data fileport-id.rrd and I have to edit and reselect the port. The issue that I can’t confirm is if the port is changing. When I select the port, I do it by device name, I start typing the device name then I see wan1, wan2, etc for the device and I select the correct interface/port I want to monitor.
When I go back to ‘fix’ the broken graph, the port is no longer selected using then name I set, it just shows a port number of 8341, so I click, type the device name and select the WAN port again (by name, I don’t see a port number) and all the data is there. At least the data doesn’t appear to be gone but for some reason the port does not stick/save.
Very strange behavior from LibreNMS, lately, I have never had these issue up until updates from earlier this year. My environment has not changed, but LibreNMS certainly has.
@dt-lib if you have a device that changes ifIndexes, you should use the ifName port association mode.
How do I do that? I’ve always added devices the same way. The graph port field I’ve always just typed the name of the device and the interfaces filter on their own.
Port association mode has been around for year and years, because bad behaving devices have been around that long. You can either set it when you add a device or when you edit the device.
I’ll take a look, thanks.
I’ve used other SNMP programs with these same devices, so far LibreNMS is the only one to lose data and break RRD graphs like this (where I have to constantly reselect the port). This sounds like a LibreNMS problem, not a device problem.
I like librenms and I’ve donated many times, but basic features are starting to fail and that’s not good.
LibreNMS is way more solid than it ever has been. I remember when devices use to break all the time and there was almost no way to tell if a pull request would break other devices.
Over the last six years our LibreNMS instance has never lost significant data aside from when either there was a outage in our virtual machine cluster – or I made some poor choices. 575 devices and almost 37000 ports in a MP-BGP/MPLS environment using only 16 vCPUs and 16 GB RAM. Currently on 26.9.0 and it’s continuing to hum along.
I’m sure it is more solid, but for me it is the opposite, I’ve had the most issues with LibreNMS in 2026 than I ever have before and I’ve had the same devices connected. The only thing that changed more, this year, than previous years are power outages. That means when those sites were offline LibreNMS is treating those devices differently (and randomly) because not all devices lose their info/graphs/etc.
However, it seems that we aren’t going to get anywhere because we get stuck in a loop of questions, so we are just stuck here each pointing fingers at the other person.
Fair, since I’m not providing as much information as is being requested, but I also don’t know what to provide since I’m just an end user and not really sure on the backend logs that are needed. It’s fine I’ll just deal with it for the time being.
Thanks.
@dt-lib I’m just trying to help you out.
Information that would be helpful:
Are there any commonalities of devices this has occurred with?
What graphs have issues? All or just some specific ones?
What OS are the devices this has occurred with?
A screenshot of what you are seeing.
What conditions surrounded the issue? Aka did the device reboot? Lose power?
You could also go to the device capture tab and run discovery and polling and pastebin the output. (Make sure to redact any private info)
What have you tried to resolve the issue?
No commonalities that I’m aware of. Everything in the field can happen randomly and independant of other cellular routers. However, the issues I am experiencing are new, the routers are not.
These are cellular routers out in the field, they can be rebooting and offline at anytime while LibreNMS has bene trying to poll.
The OS is some type of linux based OS on the routers.
The screenshot is going to show you the graph error I already provided above. It is a white box (where the graph should be) and the error I provided.
No issues that I’m aware of. The cellular routers can be out of coverage for hours at a time while LibreNMS is trying to poll, that’s where I assume the issue is occurring for all the issues in this thread (data being deleted, ports changing numbers).
I have tried nothing as I’m not sure how to troubleshoot LibreNMS that’s why I came here.