high memory load on VSX Maestro

[Arturxr]

Good afternoon, we periodically receive notifications in the monitoring system that RAM load exceeds 95%.

We have two chassis (Active/Standby), each with VSX contexts (0, 1, 2).

The output of the free -mt command on the first chassis shows the following:

Attached.

Can you tell me what could be causing the load?

The load graph from the monitoring system looks like this: (30days)

I also attached it as an attachment.

[Chris_Atkinson]

Which appliances do you have and did you max the memory on those?

Version/JHF and enabled blades info would also be helpful as a minimum....

[Arturxr]

I have a Maestro 7000. Can you tell me how to tell if I've used the maximum amount of memory?

[Chris_Atkinson]

7000 appliances can have 64G RAM populated as the max configuration.

Pending a review of the consumption & traffic volumes, ruling out leaks etc you might upgrade the amount of RAM for each appliance to resolve.

[Arturxr]

If I understand correctly, we have 32 GB of memory installed and another 32 GB is used for swap space. No access issues were observed under high load, only notifications were sent to the monitoring system

[Vincent_Bacher]

At this point, it would be particularly interesting to find out which process is consuming so much RAM. You can take a look around a bit with top or cpview.

If you use Skyline, it also provides the relevant metrics when needed, if I remember correctly. 

But since Skyline is based on cpview, cpview is sufficient if you don't use Prometheus, Dynatrace or anything else supported by Skyline.

Edith says; maybe this works and lists the processes by RAM usage but I cannot test it right now

ps -eo pid,comm,rss,vsz,%mem --sort=-rss | head

[Arturxr]

Apparently the load is related to the amount of traffic, since I don’t see any processes loading the equipment; the load is distributed evenly.

[Vincent_Bacher]

Seems so. However, top should be sorted by RAM usage rather than CPU usage. 🙂

Shift+M

edith says:

If this is related to traffic volume, I would be interested in learning more about the traffic. Is this normal traffic that one should always expect? When and how often does it occur? Why are the spikes so high that the memory load increases so much? etc.

The RAM configuration of your devices already sounds like it can handle quite a bit of traffic, and two virtual systems is a very low number.

As I said, I would be more interested in analyzing the causes in more detail than looking at what can be done with the RAM, but perhaps I misunderstood the initial post.

[Arturxr]

We discovered that one of the chassis was down, and after resolving the issues with it, the load stabilized.