Create a Post
cancel
Showing results for 
Search instead for 
Did you mean: 
HectorIvan23
Participant

SMB CPU 100%

Hi everyone,

        One of our branches has an 1800 R81.10.08 (996001683) centrally managed by a SMS R81.20 which most of time has 100% cpu usage, blades that are enabled:

  • Firewall
  • Vpn S2S
  • Antivirus
  • Identity Awareness
  • Application Control
  • Https Inspection

Within VPN we share domain services (VoIP, AD, Printing) in order to connect to other branches and main office.

We recently enabled HTTPS because some pages weren't working OK, we've been monitoring load usage on appliance and has been almost at 100% cpu usage.

Our HTTPS rulebase bypasses some categories as recommended (financial, business, etc.), internal traffic and some custom application/sites.

When monitoring with top and cpview, shows that multiple wstlsd processes are consuming most of CPU. 

Investigated further with fw ctl multik print_heavy_conn and found that most of connections were going to Office 365 services, we added a rule to accept this traffic at fw layer and still shows it the same.

At deactivating HTTPS, concurrent connections reduce in > 50% of hosts.

We tested on site and found no issues at all, 

This branch has 200 hosts average, and we want to know how can we low cpu usage and if a better equipment is required?

0 Kudos
20 Replies
Lesley
Advisor

What is the isp speed up and down. And how much in total of mb goes through this firewall you think? You can check this info with cpview command

-------
If you like this post please give a thumbs up(kudo)! 🙂
0 Kudos
HectorIvan23
Participant

Hi Lesley,

   Our ISP is 100/100, i'm not quite sure how much mb and by cpview you mean this?

 
 

 

 

 

0 Kudos
the_rock
Legend
Legend

Might be worth TAC case.

0 Kudos
HectorIvan23
Participant

Had previously an SR but was closed regarding that were too many connections for appliance capacity.

0 Kudos
the_rock
Legend
Legend

You can always open new one and reference the old case, thats what I do sometimes.

Best,

Andy

0 Kudos
Lesley
Advisor

Share please 2 screenshot from cpview -> network -> interfaces -> traffic and then RX traffic and TX

 

-------
If you like this post please give a thumbs up(kudo)! 🙂
0 Kudos
HectorIvan23
Participant

 

Done, files are attached.

0 Kudos
Lesley
Advisor

I checked but I miss column Mbps and peak (for Mbps) they are the last 2 on the right.

Looks like the screenshot just cut them off? 

-------
If you like this post please give a thumbs up(kudo)! 🙂
0 Kudos
HectorIvan23
Participant

My bad, i uploaded the files again.

Lesley
Advisor

I hope that traffic peak on LAN1 is a bug, that is a lot of traffic. 

You got anyway to verify those numbers? Maybe with monitoring tool that collected this data? 

-------
If you like this post please give a thumbs up(kudo)! 🙂
0 Kudos
HectorIvan23
Participant

I checked cpview on historical mode and found that these peaks are produced when our main ISP is down and appliance switches to auxiliary antenna (main office) and redirects all of the traffic (wan and internal)

What we don't really understand is that other appliance with 30-40% hosts and same blades enables has fewer concurrent connections (13k | 4k), so we suspect that maybe there is an infected host.

 

the_rock
Legend
Legend

That came to my mind before, it certainly makes sense. Do you see any logs indicating influx of specific IP?

Best,

Andy

0 Kudos
Lesley
Advisor

Time for a doctor spark health check:

https://sc1.checkpoint.com/documents/SMB_R81.10.X/CLI/EN/Content/Topics/drsmb.htm

I think this hardware should run with this amount of data according the datasheet:

https://resources.checkpoint.com/datasheet/check-point-1600-and-1800-security-gateways-datasheet

If you want you can share some report details here. But I would open TAC case for this. 

Could be that the report shows you have to many users / unique IP's and shows then hardware performance issues

 

-------
If you like this post please give a thumbs up(kudo)! 🙂
HectorIvan23
Participant

We previously had trouble with CCTV traffic within VPN Community, so for testing purposes we created a firewall rule dropping all traffic coming to those segments, 

On top command wstlsd processes were consuming most of CPU instances, so after installing policy, load average decreased from 10 to 0-1. Still trying to figure out if our rulebase needs an optimization or appliance has a failure.

Concurrent connections remain after those changes, but we haven't seen any significant increases on load.

Health check reports were generated and are attached to my reply.

Thanks for your help.

0 Kudos
the_rock
Legend
Legend

Hey, did you ever end up opening new TAC case and referencing old one? Personally, I would do that and also attach anything you had generated so far.

Best,

Andy

0 Kudos
Lesley
Advisor

I checked the reports and they look good. Tho it is a snapshot of the health.

Any issues at this moment? Not sure if I can help any further if the issue is gone.

Btw you do full https inspection or only: SSL Status:
Only HTTPS Categorization is enabled (Only certificate check)

-------
If you like this post please give a thumbs up(kudo)! 🙂
0 Kudos
HectorIvan23
Participant

Users were unable to access to websites and logs show only "Error" but no information further.

CPU increased and after policy install decreased but issues were still present.

Ended up opening a SR and today tech conclude following:

- Exhaustion of sockets

- DNS Issues

- Certificate Issues

- wstlsd / wsdnsd crashes for a possible bug or misconfiguration

 

We disabled blade for now and we'll be waiting for R&D and TAC update.

0 Kudos
the_rock
Legend
Legend

If you disable https inspection, all works fine? If yes, does it make any difference if blade is on and you make bunch of exceptions like below (example from my lab)

Best,

Andy

 

Screenshot_1.png

0 Kudos
HectorIvan23
Participant

Yes, navigation is OK, according to TAC https should be unabled until they keep investigated.

Our https policy bypasses internal traffic and only internet is inspected,

Regards.

0 Kudos
the_rock
Legend
Legend

To me, below looks right...

Dr. Load - Gateway Performance Report:

Number of hosts: 187
Number of connections: 12687
Connection rate: 12487 per second


Throughput:
 Receive: 8.490056 Mbps
 Transmit: 10.254888 Mbps
Packet Rate:
 Receive: 2776 packets per second
 Transmit: 3900 packets per second


SSL Status:
Only HTTPS Categorization is enabled

Blade Status:
Blade IPS is disabled
Blade AV is enabled
Blade AB is disabled
Blade TE is disabled
Blade ASPAM is disabled
VPN-RA is disabled
VPN-S2S is active

0 Kudos

Leaderboard

Epsum factorial non deposit quid pro quo hic escorol.

Upcoming Events

    CheckMates Events