Hi!
Have run into an issue with the InfluxDB integration. May be more an issue on the InfluxDB side than LibreNMS, but it affects the polling a lot.
I’ve put InfluxDB on a SSD, but it still slows down the LibreNMS polling so much that it ends up taking longer than 300 sec to finish, which causes gaps in both LibreNMS graps and InfluxDB graphs and it eventually ends up in a mess as it piles up more and more pollers.
I’ve set pollers to 2 x CPU.
I have this problem both on a large system monitoring > 2000 devices and 30000 ports and on a small system (VMs) monitoring just 50 devices.
Disabling the influx plugin immediately solves the issue.
Any hints ?
My assumption is that the tuning likely is mostly on the InfluxDB side.
/Jesper
-bash-4.2$ ./validate.php
| Component |
Version |
| LibreNMS |
1.39-11-g7e369a6 |
| DB Schema |
249 |
| PHP |
7.2.4 |
| MySQL |
5.5.56-MariaDB |
| RRDTool |
1.4.8 |
| SNMP |
NET-SNMP 5.7.2 |
====================================
[OK] Composer Version: 1.6.5
[OK] Dependencies up-to-date.
[OK] Database connection successful
[OK] Database schema correct
How much does it slow down polling? Can you give some measurements?
Hi
It ends up with all devices being listed as 999.99 seconds in The LibreNMS polling view. That time is the max LibreNMS will display…the real time is higher.
What happens is that a few polls end up taking more than 5 min and then a few more the next time and then these impact the rest, so after 4-5 polls almost all is impacted.
On my small system most devices can be polled without InfluxDB in 70s and only one device takes longer than 10 sec.
As soon as I enable InfluxDB it takes max. 15 min (3 polls) before it starts to fall apart.
On the big system where I have several SSDs it used to work when I had just a very few devices, but now with > 2000 devices it breaks immediately when I enable InfluxDB. This system also works perfectly fine wirhout InfluxDB and polls all devices in less than 300 sec.
It seems to be primarily a data inges performance issue with InfluxDB as getting data out of it via Grafana is fast.
I have the problem… If I disable influx DB integration, polling returns to normal.
All started when I changed the ip of my influx DB server…
Did u ever resolve this?
I resolved my by installing ubuntu 16.04 (I was running ubuntu 18.10)