From @szymonpk on January 27, 2017 8:5
I'm not sure how good/bad idea this is, I'll describe my issue and maybe better solution will came up during discussion.
Configuration:
- Workflow 2.10
- EC2
- TCP ELB with proxy protocol
- Router deployment with
router.deis.io/nginx.useProxyProtocol: true and proper router.deis.io/nginx.proxyRealIpCidrs.
Usually our request are running client -> elb -> workflow and it's working fine real ip is present in logs. However some domains (not even whole apps), are running through CloudFlare client -> cf -> elb -> workflow and here hell breaks loose :/ we're getting CF ip as remote address. So best option would be per domain real_ip_header, is it possible? Each domain is separate server block in nginx.conf, but is application/domain configuration available from nginx.conf template?
Maybe are there some workarounds? Switch to HTTP(S) LB is not an option.
Copied from original issue: deis/router#308
From @szymonpk on January 27, 2017 8:5
I'm not sure how good/bad idea this is, I'll describe my issue and maybe better solution will came up during discussion.
Configuration:
router.deis.io/nginx.useProxyProtocol: trueand properrouter.deis.io/nginx.proxyRealIpCidrs.Usually our request are running
client -> elb -> workflowand it's working fine real ip is present in logs. However some domains (not even whole apps), are running through CloudFlareclient -> cf -> elb -> workflowand here hell breaks loose :/ we're getting CF ip as remote address. So best option would be per domainreal_ip_header, is it possible? Each domain is separate server block in nginx.conf, but is application/domain configuration available from nginx.conf template?Maybe are there some workarounds? Switch to HTTP(S) LB is not an option.
Copied from original issue: deis/router#308