NetScaler / Citrix ADC Firewall Rules

83 thoughts on “NetScaler / Citrix ADC Firewall Rules”

1. Hello Carl,
   As always thanks for your massive insight (no pun intended…ok I’m lying). I have a point of confusion about http redirect. I have setup http redirect on NetScaler VPX 12.x.x using the loadbalancer down method. It doesn’t work . These are some newbie questions:
   1. For external connections what does my firewall have to allow? I assume TCP 80 on the IP address of the external URL?
   2. Presumably this is sent to the downed LB on the NS? That is gateway_IP:Port 80?

   If I telnet once this is done is this a legitimate way of testing and do you know what I should expect to see?

   Again I apologize for the novice questions.

   Sidebar and off topic: Do you have any posts on configuring interfaces for MPX out of the box – trunking etc, I haven’t been able to find any of yours. I need to connect a new MPX out of the box to a switch and Citrix docs arent very helpful.

   Reply

   1. 1. Port 80 to the port 80 vServer that is performing the redirect.

      See https://www.carlstalhood.com/netscaler-12-system-configuration/#portchannel

      Reply

      1. There’s a special place in virtual heaven for you. Thank you Carl for this quick response.

         Reply

2. Hi Carl,

   Im hoping you can help with this question I have. I’m looking to setup SNIP for a subnet that is behind a firewall. If I were top add a SNIP address from that subnet, do firewall ports need to be opened for the NetScaler to be able to use the SNIP address that is behind the firewall?

   Apologies, my networking experience is limited.

   Many thanks

   Reply

   1. Adding a SNIP allows you to bypass the firewall, assuming the NetScaler is connected to the subnet behind the firewall. But is this what your security team really wants? Usually bypassing firewalls is a bad security practice.

      Reply

      1. Sorry Carl – let me explain a little better – the NetScaler and it’s NSIP is infront of the firewall and the subnet would be behind it. If I assign a SNIP from that subnet, would I need certain ports open on the firewall to allow the NetScaler to use the SNIP?

         Again apologies and many thanks

         Reply

         1. Is the NetScaler connected to the SNIP subnet? What subnet is the VIP on? If VIP is on one side of the firewall, and if SNIP is on the other side of the firewall, then traffic through the VIP going out the SNIP will bypass the firewall.

            Reply

            1. Thanks for the prompt reply Carl. It is not directly connected to the SNIP subnet, but it could route to it via the firewall – I’m not sure if certain ports need to be open on the firewall for it to be able to do use the SNIP? It’s like you said – the VIP is on a different Subnet infront of the firewall and SNIP subnet is behind the firewall.

               VIP->NetScaler->Firewall->SNIP->Backend servers.

               Many Thanks

            2. You create a SNIP on a directly connected subnet. You configure a route using a router/firewall on the directly connected subnet. The SNIP communicates to the server through the router/firewall. As for firewall rules, that depends on the app and the port numbers you are load balancing.

         2. You can only add SNIPs on subnets that the NetScaler is actually connected to. But you can easily add routes for any non-connected subnet.

            Reply

3. Hi Carl,
   Do you know the communications port between the MA Agent (azure) and the NetScaler MAS OnPrem?

   Reply

4. Hey Carl, to implement remote pc access through the netscaler, do i need to open up port 80 to each client pc from the netscaler ?
   Thanks
   Gary

   Reply

   1. Port 80 is needed from the Delivery Controllers, but not from the NetScaler. Only the ICA ports are needed from NetScaler.

      Reply

      1. Cool, thanks for the prompt reply Carl, do i need to open up the ica ports between the client pc and the netscaler also ?

         Reply

         1. Yes.

            Reply

5. Hi Carl,

   I’m looking for some guidance on configuring a netscaler VPX 1000 for external access.
   Currently I have this running in a VM with 3 NICs:

   1st NIC 192.168.76.0/24
   2nd NIC 192.168.75.0/24
   3rd NIC 192.168.1.0/24

   NetScaler IP: 192.168.76.252/24 VLAN bound to 1nd NIC (0/1)
   Subnet IP: 192.168.75.251/24 VLAN bound to 2nd NIC (1/1)
   Subnet IP: 192.168.1.251/24 VLAN bound to 3rd NIC (1/2)
   NetScaler Gateway Virtual Server: 192.168.1.60/24

   From the netscaler, I can ping IP addresses on all 3 networks above as well as the router/firewall on 192.168.1.1.

   When setting up the NetScaler gateway for XenApp and XenDesktop, everything is working fine internally to 192.168.1.60/24. I’m able to telnet and open https://192.168.1.60, login to the netscaler my credentials and see/access the published apps.

   When I try to add port forwarding in my router/firewall [192.168.1.1] to 192.168.1.60/24 on port 80/443, I’m unable to access the netscaler externally on the public IP on port 80/443. Telnet to either port 80/443 isn’t working.

   If I port forward directly to my storefront server [192.168.1.25] on port 80/443, I can connect fine so I know the port forwarding rules are fine.

   Looking through various articles, I can’t see much wrong with the config. Hope you can help.

   Kind regards,

   Mark.

   Reply

   1. What is the default route (0.0.0.0)? It should be pointing to the router that can access the Internet.

      Or, you can enable Mac Based Forwarding to override the routing table for replies.

      Reply

      1. Hi Carl,

         This is exactly what the issue was. It is now resolved by creating a new default route for 0.0.0.0 to 192.168.1.1 and removing the default route for 0.0.0.0 to 192.168.75.1.

         192.168.75.1 being the IP of my Hyper-V vEthernet adapter

         What was misleading me, was the fact I could ping, connect, and resolve out to the internet. This is most likely because of the nat I setup on the 192.168.75.0/24 network.

         Many thanks for your prompt response, and thank for you all the effort you put into this site.

         Kind regards,

         Mark.

         Reply

6. what is port use for Telemetry service , After migrate from 7.8 to 7.15 PVS found console hung , restarted the SOAP service,restarted server no luck

   Reply

7. Hi Carl,

   I’d like to point out one thing in regards the firewall rule definitions for the “Insight Integration with Director” for the NetScaler MAS Firewall Rules as well as Insight Center Firewall Rules sections.

   Generally speaking, the connectivity is required from server on which Director is installed, which would commonly be separate from DDC in any mid-size to large deployments.
   As far as I know, connectivity between DDC and MAS / Insight Center is required only if Director is installed on the same machine as DDC.

   Also, it is possible to run the connectivity over HTTP, although HTTPS is recommended.

   Regards

   Reply

8. Hi Carl,

   Basic question about DNS / name resolution on Netscaler.

   I have a netscaler with two interface (Internal vs External) / Two arm mode?.
   Is it possible to send name resolution query to respective DNS server. Example InternalDomain.local should go to Internal DNS (192.168.1.1) and Externaldomain.com should go to External dns (171.168.123.122) . How?

   Regards,
   Kamal

   Reply

   1. Maybe this? https://support.citrix.com/article/CTX205898

      Reply

9. Hi Carl, Thanks for your awesome blog for the community
   I need to use SNIP for all communications (including monitor) to back end environment. Is it possible to achieve?
   1. Understand that the Netscaler uses SNIP to communicate to back end DNS, LDAP, NTP etc (if configured as LB VIP) and uses NSIP IP as source for monitor probes. To force all traffic (including monitor traffic), Is it possible to configure Net profile? If we do that, will it force all traffic through SNIP?
   2. SNIP IP can be enabled for management which means NSIP is not required to log/manage NetScaler and Putty can be enabled only for SNIP?
   3. With regards to creating Local LB VIP for LDAP, DNS, RADIUS etc inside NetScaler, Is it possible to use non routable IP as LB VIPs like 1.1.1.1 or 1.2.3.4?. This is to avoid requesting more IPs from network team

   Reply

   1. See https://support.citrix.com/article/CTX217712

      Reply

      1. 3. With regards to creating Local LB VIP for LDAP, DNS, RADIUS etc inside NetScaler, Is it possible to use non routable IP as LB VIPs like 1.1.1.1 or 1.2.3.4?. This is to avoid requesting more IPs from network team?

         Can this be done Carl or do we need to use routable IPs for LB VIPs?

         Reply

         1. Since they are essentially a loopback connection, non-routable is fine.

            Reply

            1. Hi Carl, We are experiencing issues in accessing XD VDI using IGEL thin clients. I can luanch the same VDI using our laptop. Both the laptop and IGELs are in same VLAN. IGELs are pointed to internal storefront LB. If I point the iGEL to netscaler gateway URL, it is working fine. Any thoughts

            2. Has it ever worked?

               Are you able to get Receiver logs from the Igel?

            3. Yes it was working earlier and stopped working since April and user was living with Laptop access. All our VDIs are TLS 1.2 encrypted so we are getting the generic error message as “You have chosen not to trust “QuoVadis Global SSL ICA G3”, the issuer of the server’s security certificate (SSL error 61).”. However, when I turn off SSL and it is throwing different error as “Unable to reach the xenapp server in the specified address”.

               Thanks

10. Thank you very much Carl for your prompt reply. I have one more question
    In the environment I am working on, All servers are locked with individual Windows firewall rules applied through group policy. By default, all incoming and outgoing ports are blocked with only exceptions configured through GPO.
    I can get the incoming ports to be opened (for example 80;443 on controller, 27000 on license server etc) from the article but the security team are requiring Source Ports.
    In other words, the team also need outgoing ports on servers. For example, Licensing server
    Incoming Port –
    TCP 27000
    TCP 7279
    TCP 8082-8083
    TCP 80
    Outgoing Port— need clarification
    For Example, If Controller is connecting to license server,
    Source port – Dynamic (Any port between 1025-55555) – Is it possible to lock it down to range?
    Destination port- 27000.
    After license validation when the traffic returns from license server to VDA, Will the port be reversed?
    Source Port – 27000?
    Destination port – Dynamic port?

    Reply

    1. That’s a very unusual request. There’s nothing Citrix-specific about that request. How do you do it for other firewall rules? When a browser connects to a web server on port 80, how do you limit the source ports used by the browser?

       Reply

       1. This is what I thought. I am new to the environment.
          For my understanding, On the license server, If only the below incoming ports are opened
          Incoming Port –
          TCP 27000
          TCP 7279
          TCP 8082-8083
          TCP 80
          And all the outgoing ports are blocked, Will it have any impact on licensing? Similarly for other servers/services..

          Thanks again:)

          Reply

          1. Outgoing packets from the destination machines are replies. Stateful firewalls should handle replies automatically. The destination machines do not initiate connections in the other direction, except for Controllers initiating connections to VDAs, and VDAs initiating connections to Controllers.

             Reply

             1. It is clear now Carl. Thanks for clarifying this.
                You mentioned “The destination machines do not initiate connections in the other direction, except for Controllers initiating connections to VDAs, and VDAs initiating connections to Controllers.”
                Is this also true for connection between SF and controller as well? (XML query and XML response)

             2. Correct. StoreFront sends request to Controller. Controller sends back a reply.

             3. I meant, the connection between SF and Director is also both way (XML query and response), correct?

             4. SF and Director don’t communicate with each other. But both talk to a Controller.

             5. I realised, I typed Director instead of Controller. Thanks for your answers.

11. Hi Carl,
    Thanks for the article. Really useful. I am working on a setup where Citrix MGMT servers (controllers; SF; directors) and VDA are on separate subnets and I can’t use port 80 anywhere. Either I need to use 443 or a different port. I could see in 5 places port 80 is used by default which I need to change. I have mentioned that below. Please add if I miss any
    1. From Controller to All VDAs – TCP80 For registration; I read, it is encrypted by WCF); To configure port 8080, change VDA port (8080) from VDA agent and changing on controller by using brokerservice.exe command
    2. From SF to Controller (XML) – TCP 80 (Bi) For XML brokering – To configure 443, Apply Cert on controller, Run PS command to use only 443; On SF, configure Cert; modify store to add FQDN of controller and port 443
    3. From All VDAs to Controller – TCP 80 for brokering; do I need to configure this separately? Or will step 1 ensure that this traffic also flow on 8080?
    4. From AdminPC to Controller – TCP 80 for powershell; How to configure this? Since controller is configured with cert (step 2), will this communication also goes in 443?
    5. From NS-SNIP to Controller(STA) – TCP 80 for STA tickets; How to configure this? Since controller is configured with cert (step 2), will this communication also goes in 443?

    Reply

    1. 1. That’s correct.
       2. That’s correct.
       3. Step 1 covers it
       4. BrokerService.exe /sdkport. Then I think you have to specify the port in the -AdminAddress parameter for every PowerShell command.
       5. Step 2 covers it.

       Reply

       1. Thank you very much Carl for your prompt reply. I presume for point 4, after changing the SDK port, I need to provide the new port number when launching studio (it will ask to specify delivery controller address)
          I have also seen in this blog that I got to configure /sdkport change for all other controller services (Host.exe, Monitor.exe service etc) as indicated in this https://blog.citrix24.com/xendesktop-how-to-change-used-ports/
          I will give it a try.

          Reply

12. Hi Carl,

    with NetScaler SDX 11.1-54.14, I noticed there’s a Console Access Option shown with NetScaler > Instances. Do you know which port is used here? I kicked off a tcpdump while trying to Access those VPX Console Shows only https communication. As https is opened w/ our firewalls, I can’t access the VPX Consoles though.

    Cheers,
    Jochen

    Reply

    1. It’s using WebSockets. Make sure the SVM certificate is valid.

       Reply

13. Carl,

    When creating a rule for a firewall to allow netscaler traffic, what application is using the port 7105? We are getting a ica error when opening up a session. We followed the ports needed\listed but found out that for some reason this port was not listed in the requirements.

    Reply

    1. UDP? Or TCP? If UDP, could be an Audio port.

       If you run “nstcpdump.sh port 7105” on the NetScaler, do you see it sending that port?

       Reply

       1. Thank you for the response.

          We had our Boundary protection team watching the traffic and gathering the data. From what we have seen in the data, that port is allowed now. But we still receive the error.

          We have users from other locations that are able to use the Netscaler with no problems. What we are thinking is that at some point our Boundary team removed the rule allowing this site access due to lack of use. The site in question is our backup site. The rules were not supposed to be changed or removed.

          What I am going to ask our team to do is compare the FW rules between the sites and the proxy server as well to ensure that they are set the same.

          Reply

14. Hi Carl, please add 54321-54323 from target device to PVS Servers console ports, SOAP Service, used by Imaging Wizards.
    Thanks for all

    Reply

    1. Thanks for notifying me.

       Reply

15. Hi All, I have setup netscaler 11.1 vpx on AWS and everything is fine but when launching applications it doesn happen. From directly storefront its working fine. I just came to know that 2598/1494 is getting reset itself by delivery controller. Every ports are allowed but still these two ports are getting reset itself. Please suggest if you have any solutions.

    Reply

16. Hi Carl,

    We are using Netscaler MPX5500 in our citrix environment. Our environment is secure through SSL VPN and WAF. Client . Client wants to present the StoreFront directly out to the internet and and therefore relieve ourselves of the dependency on the NetScalers.What would we gain and what would we lose? My concern here is how we secure our environment without netscaler ? How we do the encryption to secure https connections without netscaler.

    Reply

17. Hi Carl

    Thanks for all information.

    In our environment we are able to telnet( on 3009, 3010, 3011 and 22) from Site A to Site B but vice versa is not happening for GSLB setup. As per Network guy GSLB services are not running on Site A as they are unable to telnet from FW(in btw SiteA and SiteB) to SiteA. Whereas same is happening from FW to SiteB. However we have installed the GSLB service properly while configuring. I have also tested to telnet on self GSLB Site A IP via same GSLB ports and fails which indicates some issues in GSLB services in Site 1 but unable guess where it could be. Please can you help me with a hint or possible configuration to check?

    Thanks in advance.

    Reply

18. Hi carl, What is the difference between Local GSLB Site IP SNIP and SNIP? Firewall ports mentioned in this blog are for SNIP? I have a requirement to setup GSLB.

    Reply

    1. Have you seen this yet? https://www.carlstalhood.com/global-server-load-balancing-gslb-netscaler-11-1/#planning

       Reply

19. Hey Carl,

    This was GREAT help for me.

    Quick question though, I have a LAB with a 3 legged scenario: 1 Subnet for Management (NSIPs), One subnet for DMZ, and another Subnet for backend services (LAN).

    In this case, since I am isolating management, I notice that the source for the perl scripts is the SNIP, not the NSIPs. Is this normal behavior?

    Reply

    1. If there is no direct route, it will use the SNIP. But I’m not sure if it changes the source IP.

       Reply

20. Hi Carls,
    is there any ports to be opened between NSIP and SNIP.
    If the two ips are in different subnets.

    Reply

    1. No. The NetScaler can communicate between those IPs from inside the appliance.

       Reply

21. Hey Carl, thanks for the Information.

    I have one Questions for NetScaler VPN.

    Which Firewall Ports are needed for the VPN Setup? My NetScaler is in DMZ with a VPN vServer. Is only Port 443 to my StoreFront from my SNIP needed? Because I think “Any” from my SNIP to my LAN cannot be a Resolution…

    Thanks an best Regards
    Mark

    Reply

    1. What traffic is going across the VPN tunnel?

       If you aren’t doing Intranet IPs, then everything comes from the SNIP and SNIP needs access to everything the users need to access.

       If you are doing Intranet IPs, then you open firewall from the Intrnaet IP to the whatever the users need to access.

       Reply

22. what about option 66 on the DHCP server? shouldn’t that be on this list?

    Reply

    1. Are you asking for a firewall rule if you’re using a different TFTP server than the one installed on PvS?

       Reply

23. cannot rollback the fw rule now…customer has strict change mgmt for that..(read “the process to heavy so will leave it there for now) but this must be tested elsewhere

    Reply

24. No it was actually OFF for some reason….my bad

    Reply

    1. Enabling it removed the firewall requirement?

       Reply

25. Access from StoreFront nodes version 3.6 to NS LB VIP needs to be open on port 443 and https.
    Found out this the hard way…it seems the SF nodes need access to /discover url. I am not sure this has to do with the new 3.6 feature “no need for hostfile modification” stuff but worth mentioning maybe in the FW rules

    Reply

    1. Is StoreFront configured with Loopback set to OnUsingHttp?

       Reply

26. Hi Carl,
    great article! But I think there is something missing in the PVS section.

    You wrote:

    TargetDevices -> Provisioning Servers
    UDP 69 – TFTP
    UDP 4011 – PXE
    UDP 6890-6969 – Streaming

    But shouldn’t it be more like this:

    TargetDevices -> Provisioning Servers
    UDP 69 – TFTP
    UDP 4011/67 – PXE/Broadcast
    UDP 6910 – Target Device logon at PVS
    UDP 6910-6930 – streaming service (default with 8 threads per port)
    UDP 6969 – Two Stage Boot (If ISO or USB is used)

    And also I’m missing the PVS to PVS communication:

    UDP 6890-6909 – PVS Inter-Server communication

    Please correct me if I’m wrong

    Best Regards,
    Sebastian

    Reply

    1. Isn’t 67 only needed for DHCP on PvS? If DHCP is separate from PvS, then isn’t it 4011?

       6890-6969 should encompass all of the ports. I always increase the default TD ports from 6910-6968. But if 6890-6909 is only used between servers then I could clarify that.

       Reply

       1. Hi Carl,
          actually it’s the other way round.
          Port 4011 will be used if PXE is on the same machine as DHCP. And port 67 is used if it’s separated (PXE Broadcast). I just added port 67 explicit for the sake of completeness. 🙂

          And yes, 6890-6909 is only used for inter-pvs communication.
          Didn’t notice that you wanted to point out the reconfiguration for the streaming ports – sorry!.
          But you’re right – it’s a good thing to do! Maybe (to prevent misconfiguration) you should point that the range has to be manually extended and otherwise the configured ports won’t be used.

          Best Regards,
          Sebastian

          Reply

          1. Updated. Thanks for noticing.

             Reply

27. Hi Carl,

    Can we have LDAP and XML service servers in different subnet, from SNIP? I am able to ping the Domain Controller and CITRIX Controller Servers from the NetScaler, however I believe that goes through the NetScaler IP.

    If yes, how can we configure the communication between SNIP to LDAP, DNS & XML Service? Using Gateway Routes?

    Regards,
    Swapnil

    Reply

    1. If the NetScaler is not connect to the same subnet as the back-end servers then NetScaler will send the packets through a router. If you only have one connected interface then it will go through the default gateway. If you have multiple subnets then you need to configure the routing table correctly.

       Reply

28. Hi Carl, thanks for the article.
    What would be the required ports to acces the SVM GUI from and the administrator´s machine?, and the same to the Xenserver IP?
    And also, does the Netscaler GUI versión 11 still requieres the java ports?

    Reply

    1. You would want 22, 80, and 443 to access SVM and XenServer.

       In 11 and newer, Java is not needed from the administrator machine. But still needed in 10.5 build 56 and older.

       Reply

29. Hello CArl.

    Thanks for article. I need a help for NS. Netscaler MPX appliiance version 11 or version 10.5.6 can configure as a layer 4 firewall. So i need a link or document from citrix website that Netscaler ‘s certfification approved by global authorities? Thanks for help.

    Reply

    1. I don’t think NetScaler is intended as a L4 firewall. It has ACLS and other security features but that’s not the purpose of the appliance. I always put firewalls in front my NetScalers.

       Reply

30. If the SQL server instance is not default named..servers use UDP1434 to connect to database

    Reply

    1. Added. Thanks.

       Reply

31. hi,
    which source IP (on the netscaler) and target port are used for a CERT (smartcard) authentication server policy ?

    thanks

    Reply

    1. A CERT policy should be looking at the contents of the smart care certificate to retrieve the username. I don’t think it communicates with anything.

       The SSL vServer would have Client Certificates enabled. This compares the client certificate signature with a CA certificate that is bound to the SSL vServer. Optionally, you can configure CRL checking (direct or through OCSP) that would require communication with external servers. I’m guessing it uses the SNIP but I’m not sure. If you are able to set this up in a lab, run nstcpdump.sh on the NetScaler to see which IP it is using for CRL checking.

       Reply

       1. yes you’re right, i have just discovered the same thing. THanks for your quick reply !

          Reply

32. Hi,
    very good article, I think that DNS by default use NSIP (it’s like the authentication flow). Netscaler uses SNIP only in case of LB internal rules….

    Reply

    1. To verify the source IP, SSH to NetScaler, run shell, run nstcpdump.sh port 53. Do something on NetScaler to cause a DNS query and you’ll see the Source IP.

       Reply