Alert checker for high memory usage on Linux
I'm trying to setup a high memory usage alert on Ubuntu Linux hosts. I'm using type entity mempool_perc in alert test 'mempool_perc greater 90'. However mempool_perc reports all memory used, including cache/buffer memory used by Linux, but still available to be released for applications as necessary. So my alert checker generates false alerts as soon as memory usage including cache/buffer exceeds 90%, which it does quit often on most of my servers. How can I setup a memory alert checker for available physical memory below a certain threshold?
Observium online guide states: 'On Unix/Linux systems we generate a special "Physical Memory" memory pool entity by calculating memory usage minus buffer and cache. Use this for alerting on Unix/Linux' But not really sure how to use that entity? Is the entity name 'Physical_Memory', does it return result in bytes, percentage ?
Thanks in advance
Med venlig hilsen | With kind regards | Mit freundlichen Grüßen Ulrik Paludan | IT Operations NOVASOL AS Virumgårdsvej 27 - DK 2830 Virum Phone +45 39 14 31 82 ulrik.paludan@novasol.commailto:ulrik.paludan@novasol.com
www.novasol.com | www.dansommer.dkhttp://www.dansommer.dk/ | www.cuendet.comhttp://www.cuendet.dk/
NOVASOL AS IS A WYNDHAM RENTALS COMPANY
Please consider the environment before printing this e-mail
"Physical memory" is the mempool_descr of the memory pool you want to match on UNIX systems.
Note that you seem to misunderstand the scope of entities and their attributes and metrics. A device doesn't have a mempool_perc, a device has an number of memory pools each with their own mempool_perc.
On UNIX systems the values given by the device are mostly worthless for alerting by themselves, so we generate a "fake" memory pool by doing some maths.
It's this "fake" pool you want your memory alert to match, by using the association rules.
Make sure you're running the latest version, because association rules have changed a lot!
adam.
On 2018-07-11 01:37, Paludan, Ulrik wrote:
I'm trying to setup a high memory usage alert on Ubuntu Linux hosts. I'm using type entity mempool_perc in alert test 'mempool_perc greater 90'. However mempool_perc reports all memory used, including cache/buffer memory used by Linux, but still available to be released for applications as necessary. So my alert checker generates false alerts as soon as memory usage including cache/buffer exceeds 90%, which it does quit often on most of my servers.
How can I setup a memory alert checker for available physical memory below a certain threshold?
Observium online guide states:
'On Unix/Linux systems we generate a special "Physical Memory" memory pool entity by calculating memory usage minus buffer and cache. Use this for alerting on Unix/Linux'
But not really sure how to use that entity? Is the entity name 'Physical_Memory', does it return result in bytes, percentage ?
Thanks in advance
MED VENLIG HILSEN | WITH KIND REGARDS | MIT FREUNDLICHEN GRÜßEN
ULRIK PALUDAN | IT OPERATIONS
NOVASOL AS
Virumgårdsvej 27 - DK 2830 Virum
Phone +45 39 14 31 82
ulrik.paludan@novasol.com
www.novasol.com [1] | www.dansommer.dk [2] | www.cuendet.com [3]
NOVASOL AS IS A WYNDHAM RENTALS COMPANY
Please consider the environment before printing this e-mail
Links:
[1] http://postman.memetic.org/www.novasol.com [2] http://www.dansommer.dk/ [3] http://www.cuendet.dk/ _______________________________________________ observium mailing list observium@observium.org http://postman.memetic.org/cgi-bin/mailman/listinfo/observium
participants (2)
-
Adam Armstrong -
Paludan, Ulrik