Where to get "wire" memory usage

[cengjingcaihua]

I was trying to figure out a way to calculate the memory used by services and show it in Grafana. From TrueNAS's websocket messages I can see the fomula is "services = active + inactive + laundry + wire" but in graphite's measurements I can only get "wired" instead of "wire". Where is TrueNAS getting this number?

 

[sretalla]

I guess it's a typo/abbreviation somehow... the correct term is "wired memory".

 

[cengjingcaihua]

Sorry for the late reply. Just lost the entire pool due to a bad PSU that cooked all my hard drives.

I guess it's a typo/abbreviation somehow... the correct term is "wired memory".

It's not a typo. If you take a look at the websocket, the number of "wire" is much smaller than "wired".

 

[jgreco]

Wired memory generally refers to memory which is required for the operation of the system, and is effectively locked ("wired down") so as to force it to remain resident and avoid getting swapped to disk. Different systems have slightly different definitions of this, but classic examples would be kernel stuff such as buffers.

I would note that this doesn't seem to fit into the definition of "services" and therefore it seems unlikely that it would be a component of services memory consumption.

bhyve has its "-S" option (wire guest memory) to prevent swapping of VM guest memory by the host; it might be reasonable that this would be added to "services" if we assume that VM's qualify as services. It's probably stupid that these two quantities differ by a single letter.

Just an educated guess. If you want to know for sure, UTSL.

 

[Juan Manuel Palacios]

I was trying to figure out a way to calculate the memory used by services and show it in Grafana. From TrueNAS's websocket messages I can see the fomula is "services = active + inactive + laundry + wire" but in graphite's measurements I can only get "wired" instead of "wire". Where is TrueNAS getting this number?

Have you managed to solve this mystery? Do you have a number in Grafana that resembles what the TrueNAS dashboard is showing?

And… would you mind sharing the details? ;) (e.g. the query in Grafana?)

Thanks in advance!

PS: I'm a Grafana and InfluxDB neophyte, so please bear with me!

 

[cengjingcaihua]

Have you managed to solve this mystery?

No, I couldn't find it anywhere. The metrics I have in the dashboard are CPU load, temps, network, ARC hit/miss and disk related measurements.

 

[Juan Manuel Palacios]

No, I couldn't find it anywhere. The metrics I have in the dashboard are CPU load, temps, network, ARC hit/miss and disk related measurements.

OK, thanks for the info, though.

Question, though, if I may: what's the query for ARC hit/miss? And on what type of widget do you plot it?

Thanks!

 

[cengjingcaihua]

Question, though, if I may: what's the query for ARC hit/miss? And on what type of widget do you plot it?

Code:

SELECT last("hits") FROM "arcstat_ratio_arc" WHERE $timeFilter GROUP BY time($__interval) fill(null)

Take a look at this table: arcstat_ratio_arc

 

[Juan Manuel Palacios]

Code:

SELECT last("hits") FROM "arcstat_ratio_arc" WHERE $timeFilter GROUP BY time($__interval) fill(null)

Take a look at this table: arcstat_ratio_arc

Thank you for that information!

But, a couple of questions: what type of widget are you plotting it on? And were you going to link to something when you said "take a look at this table"?

Thanks!

 

[cengjingcaihua]

I'm exclusively using the time series widget.

Have you managed to solve this mystery?

I am using c_max + 1G - free - arc to calculate the services memory usage. Not a graceful way to solve this mystery but it works.