the iRule below can be used to inject a TLS SNI extension to the server side based on e.g. HOST-Header values. The iRule is usefull if your pool servers depending on valid SNI records and you don't want to configure dedicated Server SSL Profiles for each single web application.
Nice irule... I used this irule for APM portal access resource by replacing the HTTP_REQUEST event by ACCESS_ACL_ALLOWED.
and it worked.
I've slightly optimized the iRule. Make sure to update your toolbox... ;-)
4 lines for such need.... with encoding explanations!
And its now ~4-times faster than before :-)
Impressive, five stars! :)
with this optimized code, do you still use the "Client side to server side SNI relay iRule" or do you always use this one?
I was redirected from this thread (https://devcentral.f5.com/questions/help-with-sni-not-being-passed-to-pool-servers#comment50962) to this one by Stanislas. I was struggling with two certs/two profiles with SNI support. couldn't make it work. I just tested this irule and it works fine. I wanted to use the default config instead of an irule as I already have a complex irule/apm setup but this solves my issue for the moment. I had a question in mind about how to make it work with two certs but I answered my question with a quick test. I created a client profile (no sni setting defined) and added the two certs ( with only one key otherwise you get the error 'Client SSL profile cannot contain more than one set of same certificate/key type'). I created a server profile (no sni no cert just inherited from th default ssl server). Pluged the iRule to the VS and voilà ! :-)
But I'd like to know if configuring multiple profiles with sni entries (and a default sni profile) on a single VS work !
you can configure on the virtual server multiple clientssl profile (to manage client servername certificate selection) and with only one serverssl profile with this irule.
I still can't make it work. For some reason I don't know, the F5 is still sending its own hostname in the Client Hello.
@angorya f5 can't know its own host name in irule. I guess you mean the host name of the virtual server!
If you use APM With rewrite mode, read my first comment!
This might be a bit out of context, but you can actually extract the hostname of the F5 from an iRule:
Stanislas, you're correct, I expressed myself incorrectly. It's sending the VS hostname.
I'm using APM with rewrite mode, so I followed your suggestion when I read this thread (I did test Kai's iRule too).
I had the iRule log what it does, so I can see that the hostname that it's going to inject is the correct one, but in the packet capture I still see the VS hostname.
Edit: Just in case, this is in 13.0.0.
@angorya, I am seeing the same behavior you are. I have a portal access profile setup to route traffic to an internal app. I used the rule above and the tcpdumps still show the original server name instead of the injected server name.
Patrik mentioned: tcl_platform(machine)
Patrik mentioned: tcl_platform(machine)
Hi Patrik, instead of using the slightly bugged tcl_platform(machine) global variable, you may try the build-in TCL command [info hostname] to easily access the HOST-Name of your F5.
I am struggling the same as Angorya and Dane to get this working with Portal-Access. Replaced HTTP_Request with ACCESS_ACL_ALLOWED and its still sending the original SNI
F5 has it as a Known Issue for 13.0:
653495-1 : Incorrect SNI hostname attached to serverside connections
Component: Local Traffic Manager
SNI hostname submitted to a virtual server on the client side is sent to server side, even if there is a different hostname specified in the server SSL profile.
-- Client side ClientHello contains SNI.
SNI is sent from client to server without stripping or rewriting the SNI.
It seems we have to wait for them to patch it...
Seems to be an dead end then, since you can not remove an existing SNI extension from the Handshake for now (see my prequisites Step 3.).
The last chance I see for you is by utilizing a VIP-targeting-VIP setup, so that you can still use SNI to flip SSL Profiles for the client side connection (aka. external VS) while making sure to not forward a SNI value to APM (aka. internal VS). At least this approach will allow you to mitigate the listed condition "Client side ClientHello contains SNI." and would then allow you to insert a SNI value of your choice on the connection between the internal VS and backend servers.
Tested on 11.5.4 and works!
I have two virtual servers with multiple server ssl profiles applied. One is working, the second only work with one domain. Replacing the multiple server ssl profiles with this irule solved the problem. Thanks!
You're both very welcome!
could you please explain, how to tweak the sni_value? I'm to stupid to get this...
we've 2 servernames behind a Portal-Access (different from the FQDN of Client-Side), now only the one works, I insert the servername in the sni-field of the SSL-Profile.
But what would there be the $sni_value ? At the moment I don't get it.
look at the first comment above...
the iRule uses the HTTP_REQUEST event to extract the HOST-header value of the client requested ressource. This value is then used as Server_SSL SNI-Value.
If the client requests https://www.domain.com/ the Server_SSL SNI-Value will become www.domain.com
If the client requests https://ww2.domain.com/ the Server_SSL SNI-Value will become ww2.domain.com
Note: You may use whatever logic during the HTTP_REQUEST event you like to decide which SNI-Value should be used to negotiate the Server Side SSL connection.
Important: To make this iRule work you have to make sure that you dont specify a SNI-Value in your Server_SSL_Profile. This will allow the iRule to simply inject the SNI-Value without clearing existing values (aka. clearing existing values is not supported). If you set the "Server Name" option in your Server_SSL_Profile, the SSL negotiation will most likely fail...
that did it, thanks a lot.
Hi Kai, thanks too for explaining and for the Rule