Spark Embedded physical memory high

[dkzndkqh]

Currently, my SMB(1550, R81.10.15) appliance is using too much memory. However, even when I check with top, I can't figure out which process is causing the high memory usage.

Could you please let me know what might be causing it?

[Tom_Hinoue]

I strongly recommend you to upgrade to the latest R81.10.17 which includes a lot of fixes including performance optimization for the 15X0 appliances (2GB RAM models). Try checking the usage after upgrade, and if you still experience high memory, you might want to engage with TAC for further troubleshoot.

p.s.
I hope someone can kindly move this to the Spark (SMB) section 🙂

[dkzndkqh]

Thank you for letting me know! By the way, is there a big difference before and after the upgrade? I feel like I should keep the upgrade as a last resort.

[Chris_Atkinson]

You might also wish to explore the script in sk183290 - Optimizer Script for Quantum Spark Gateways 1500 series .

[dkzndkqh]

Thank you for the information. I have a few questions. In the SK document, it says to disable features like IoT, SD-WAN, Smart Accel, IoT Statistics, Device information logging, Hit Count, Privacy Settings, Logging of allowed traffic, Probing in Internet connections (depending on the script), and Batch recognition to decrease CPU and memory usage.
Does this mean that disabling these features will reduce resource consumption?

[G_W_Albrecht]

Yes.

[dkzndkqh]

We current don’t use  IoT, SD-WAN, IoT Statistics blades but the others can’t disable

[G_W_Albrecht]

e.g. Hit Count is not relevant here, i would think. I suggest to run the script an then enable the absolutely needed features again...

[dkzndkqh]

Since we're currently using Smart-1 Cloud, there are limitations on what we can optimize directly on the appliance.

[G_W_Albrecht · Accepted solution]

No worry ! Let us look at the .lua Script in detail to learn more about the performed steps ! The script first checks for Smart-1 Cloud / Central Management:

printWarn("Centrally-managed mode. Skipping the disabling of IoT, SD-WAN, SmartAccel, Hit Count, and Privacy Settings")

All these steps are left out - so the SK is somewhat misleading ! So next, in script gw_optimizer_V12.lua probing will be disabled - you need to use the script gw_optimizer_V12_no_probing.lua so probing settings will be unchanged ! Next step sounds good.

increase sfwd RAM for 1600 and later 

 Sounds good ! Then:

disable batchReognition.lua script

and restart the main process:

sfwd restart

Now for a dessert:

create bg script for pfrm cleaning (and a correspoinding script to stop it)

This will add cleaning to userScript.

printSucc("gw_optimizer_bg.sh was created to periodically clean the /pfrm2.0 directory (run '/storage/gw_optimizer/gw_optimizer_stop.sh' to stop it)")

So you can use this without fear i think...

[dkzndkqh]

Sorry for the late reply, and thank you so much for your help. I’ll keep this in mind when I work on it later!

[the_rock]

Did you try run below?

Andy

https://sc1.checkpoint.com/documents/SMB_R81.10.X/AdminGuides_Locally_Managed/EN/Content/Topics/DrSp...

[dkzndkqh]

Are you perhaps talking about installing Dr. Spark report?

[the_rock]

Correct.

[dkzndkqh]

I downloaded the report and reviewed it, and it also shows that the memory usage is high

[the_rock]

I would open TAC case and have them review it as well.

Andy

[dkzndkqh]

thank you for you help

[the_rock]

As the guys suggested, maybe upgrade to latest version, hope that helps. 

Andy

[Chris_Atkinson]

When you review sk183153 R81.10.17 for Quantum Spark Appliances you'll see some similarities to the script shared with respect to optimizations.

Beyond that you may need to work with TAC or your SE if there are specific sizing concerns that persist.