This is probably caused by the counters used by the detailed graph not correctly handling the very large numbers.
These values come from UCD-MIB, quite old and probably not designed for huge values.
Adam.
Sent with AquaMail for Android http://www.aqua-mail.com
On 14 August 2015 11:04:20 am "Louis Bailleul" louis.bailleul@phangos.fr wrote:
Hi,
Just discovered a weirdness when I added a server that has 3TB of RAM : the ucd_memory graph is totally messed up. It only displays negative amount of Cache/Shared buffers.
The legend shows that it didn't handle the 3TB correctly. But it seems to be fine for the health/memory view where the correct amount is displayed everywhere.
I have attached the debug output of the poller for this device. Best regards, Louis
observium mailing list observium@observium.org http://postman.memetic.org/cgi-bin/mailman/listinfo/observium