High memory consumption in SMS.

[Matlu]

Hello,

Currently we are having a high memory consumption in the SMS (It is what can be seen visually from the SmartConsole), and when we validate with the command "top", to see the processes, as I can recognize, it is the process "java" who is at the moment, consuming a lot of memory.

Is this an expected behavior?

Is there a way to "kill the process" that is consuming more memory?

I have a memory consumption of my SMS that is at times over 70%.

I keep version R81.10
SmartConsole 81.10.9600.414

Thanks for any comments that can help me

[the_rock]

Yes, you can run ps -auxw, find process id and then run kill -9 processid

So say process id is 50000, you would run kill -9 50000

By the way, if java is consuming high memory, that could be database issue, so if reboot of mgmt does not help, I would check below

Andy

https://support.checkpoint.com/results/sk/sk123417

[just13pro]

Is it safe or is it recommended to kill the process?

[the_rock]

Well, personally, I would say its safe, done it many times. Is it recommended? That might be more TAC question. Again, my own opinion, I recommended it many times, its least intrusive, unless you are able to cpstop; cpstart or reboot.

Andy

[Chris_Atkinson]

Which JHF and what Management blades are utilized/active, how many gateways is it responsible for?

For production use 8GB RAM is the _minimum_ recommend per the release notes.

[Matlu]

Hello,

SMS as such, does not have "extra" blades enabled.

It is important to mention that an increase in the memory of the SMS OPEN SERVER was observed when upgrading from R81 to R81.10 version.
Can the upgrade, from one version to another, cause this behavior?

I have Take 109 -> R81.10

This is the output of the command "ps -auxw".

This issue of knowing whether to "kill or not" a process that we detect that is consuming enough memory, it is better to revalidate this with the TAC?

Greetings.

[the_rock]

Hey,

Were you able to sort this out?

Andy

[Matlu]

Hello, my friend.

The memory seems to be "relaxing".

It is no longer shooting "at every instant" the value of its consumption.

It seems to be less "stressed", HAHAHAHA.

A query, as a "recommendation", what is the value of the percentage of memory, where I "should" feel concerned about my GW and / or SMS.

Is it perhaps to take into account that I do not exceed a certain threshold of consumption?

More than 70%, it is already to be "afraid" of the equipment? 😕

Greetings.

[Chris_Atkinson]

The minimum hardware requirements for management is 8GB memory, start by meeting or exceeding this as your baseline and monitor for swap utilization etc

[the_rock]

All those numbers are relative. Personally, I would say 60% is my limit, but again, it depends on so many factors.

Andy

[PhoneBoy]

You'll notice that some of the Java processes have a priority of 39.
This is the LOWEST priority, which means if other processes need CPU, they'll get it.
Java is used for the API server and indexing logs.
Killing these processes without first consulting with TAC is not recommended.

[Hugo_vd_Kooij]

Within top use the "c"button to toggle the commandline shown in top. It will be a bit more clearer on what process it actually is.

Well.Linux tends to use memory when it is available to speed up things. So any longer running system still below 80% memory usage might still be running fine. (unused memory is wasted memory.)

So at this point I don't really see a problem. In my view any Check Point system which starts to use swap memory is having a serious issue. But using  most of the memory isn't a problem in itself.

[the_rock]

Thats a good point Hugo, but in my view, if memory was not an issue before, it definitely could be a problem now.

Just my 2 cents.

Andy

[Bob_Zimmerman]

You're not hitting a memory limitation. Even with only 6 GB of RAM for the management, you aren't swapping.

That Java process is almost guaranteed to be CPM. Take note of the PID and compare it to the PID of CPM in the output of 'cpwd_admin list'. CPM fundamental to the operation of the management server, so killing it would effectively stop the management server from working.

25.5% of 6 GB is only about 1.5 GB. On one of my lab boxes with very few objects beyond the default set, CPM process is consuming a bit over 1 GB of RAM. Your consumption seems pretty normal for a management server which is actually used by people to manage real firewalls.