Does Membase (Couchbase) admin statistics have a metric for memory hits vs disk hits?

[Orit Alul]

Hi,

I'd like to measure the memory hits ratio (related to the total amount of hits memory + disk).

Can you please recommend what would be the best way to do that?

Can you please explain what the following metrics measure?

hit_ratio

get_hits

get_misses (is it a cluster miss (key not found) or a cache miss?)


Thanks,
Orit

[Perry Krug]

Orit, the Membase UI presents this information as the graph labeled "Cache miss ratio".

To determine it yourself, you will need to calculate it with ep_bg_fetched (how many items have been fetched from disk) versus get_hits (how many items were serviced by the system overall).

-'hit_ratio' is not something the server provides directly, it is usually computed by an external system by measuring get_hits versus cmd_gets but this refers to the hit_ratio of the cluster (key not found) as opposed to coming off of disk.

-'get_hits' are the same as memcached and refer to how many gets were serviced with a key that the system had

-'get_misses' are also the same as memcached and refer to how many gets were serviced with a "key not found".

Perry

Perry Krug Solutions Architect direct: 831-824-4123 email: pe...@couchbase.com

[Anton Mostovoy]

Perry, could you please clarify why "'hit_ratio' is not something the server provides directly, it is usually computed by an external system by measuring get_hits versus cmd_gets". Is it just about possible timeouts that could cause the client to see more misses?

So, is the existing hit_ratio metric calculated by get_hits/(gets_hits+get_misses)?

Thanks!