Hey Scott,

Assuming you are using two-way TLS with client certificates for authentication, 
I recommend configuring your ELB for TCP passthrough so that the TLS handshake 
is between the end-client and the NiFi Registry Server (in other words, no 
decryption/termination of the TLS connection happens in the ELB). If you are 
using some other form of authentication (e.g., LDAP), you will need to 
configure your ELB to trust the self-signed key NiFi Registry is using. I'm not 
sure how to do that as I've never run an ELB with that configuration before.

Also, just a note about using an ELB with NiFi Registry:

NiFi Registry is currently only supports single-instance use as persisted data 
and in-memory state is not synced between multiple instances. Are you hoping to 
use the ELB for actual load balancing, or is it just to take advantage of other 
ELB features, such as forwarding and security group rules? If the plan is to 
load balance multiple Registry instances, just be aware that you will probably 
run into some unexpected behavior. (As you mentioned using authorization, that 
is one case where I know the in-memory cache of the persisted data will not 
refresh across instances, so even if you were using some sort of shared network 
file system attached to multiple Registry instances, such as EFS, it would not 
work the way you hope.)

Hope this helps,
Kevin

On 3/19/18, 10:20, "Scott Howell" <[email protected]> wrote:

    Thanks for the quick response.
    
    A couple of things I am seeing.
    
    1. There is no error, I don’t see anything in the logs once the service 
comes up. This is because the health check is not even hitting the instance 
when secure. 
    
    2. Nothing interesting in the nifi-registry-app.logs. That was my concern 
because on my nifi instance I can see the health check hitting the instance 
from the ELB. This does not happen on the nifi-registry instance.  I see the 
service startup and it tells me what domain and port I can access the UI but 
nothing else after that.
    
    3. When I am on an instances in the same private subnet I am able to curl 
to the instance I get the TLS SSL which tells me the keystore is on the server. 
I am using a JKS keystore that is self-signed by the company I work for.
    
    > On Mar 19, 2018, at 9:10 AM, Bryan Bende <[email protected]> wrote:
    > 
    > Hello,
    > 
    > What error are you getting when you cannot access the UI?
    > 
    > Is there anything interesting in nifi-registry-app.log regarding
    > authentication/authorization when this happens?
    > 
    > Can you access the UI securely without going through the ELB?
    > 
    > Thanks,
    > 
    > Bryan
    > 
    > 
    > On Mon, Mar 19, 2018 at 10:05 AM, Scott Howell <[email protected]> 
wrote:
    >> I was able to stand up nifi-registry behind an AWS ELB non-secure. 
Everything was working great and was able to access the UI anonymously. I set 
up the authorization just like on my nifi instances along with the authorizers 
and identity-provider. The service comes up without errors and everything looks 
good but the health check does not pass and I cannot access the UI to login. I 
was wondering if anyone else has ran into this issue using nifi-registry.
    
    


Reply via email to