Website is slow when put behind vm-series 300

[Tariq87]

We have deployed vm-series 300 in AWS recently and put our production site behind it, but we are seeing a performance degradation, the website is taking around 2-3 mins to load for the first time which normally it didnt take, we have not put any url filtering profiles yet but yes we do have some security and nat profiles in place(which normal I believe), I just wanted to understand what could be the reason for the slowness and what all things can be checked in PA to troubleshoot.

PS: I am attaching a screenshot of developers tool console to show time it takes for the page to load first time

[rapatil]

Hi Tariq87, 

Can you please open a support case so we can look into this issue and help to troubleshoot it further?

Thank you.

[rapatil]

Hi Tariq87,
Can you please open a support case so we can look into this issue and help
to troubleshoot it further?

[jmeurer]

How are you directing traffic to the firewalls?  If you are using an ALB/NLB, ensure that your routing allows the firewalls to communicate with all of the LB Subnets via the same interface that is specified in the Target Group.   Similar question on the backend.  If the application is deployed across multiple subnets, ensure the firewall can route to all of the backend subnets via the SNAT interface.

[Tariq87]

Yes, we are using NLB, which is deployed in 2 subnets across 2 az's, the backed firewalls are also in the same AZ's. The backend to the firewall is an Haproxy which are also spread across the same az's.

What else can I check?

[jmeurer]

There is a very good chance that your external interface cannot route to the NLB subnet in the other zone or that your Trust side interface cannot route to the other zone.  If you can post your subnet lists and your firewall VR "More Runtime Stats", we may be able to see the issue.  

Also watch that you do not have "Automatically create default route pointing to default gateway provided by server" enabled on both of your interfaces.

[Tariq87]

I am sure of the second point you have mentioned, no we did not do that.

For the first, I'll check and get back to u

[Tariq87]

How can I do that?

[tostern]

@Tariq87 

see below on the picture. You have to make sure that on the interface on the firewall are you not select "Automatically create default route....."

 

[Tariq87]

Yes, we have not selected this for the trust interface

[jmeurer]

Please see me earlier note about posting your resultant route table from the firewall and your subnet layout.

[Tariq87]

10.163.6. and 10.163.5 are untrust interfaces(subnets) and 10.163.11 and 10.163.15 are trust interfaces

[jmeurer]

What subnets are your LB and back in?  The route that I question is the /22 on only one of the firewalls.  If you have backend traffic that is not in the same subnet as an interface, you could be routing it out Eth1/1 but snatting it to Eth1/2.  That will cause a timeout.  

On the flip side, you have a 10/8 route via eth1/2, this is going to prevent your LB health checks from return to the cross zone subnet via eth1/1.

You should have 0/0 via eth1/1 and static route for the NLB subnets also via eth1/1.  You can then use the 10.8 route to get to all other subnets via eth1/2.

[Tariq87]

i need some time to verify this, need to look into it tomorrow morning and will get back to you

[Tariq87]

We have found out that it is the decryption profile which is causing the slowness, I have opened a ticket with PA support