Red Hat Bugzilla – Bug 996936
[origin_runtime_209]Traffics don't go to other haproxy instances except for head gear for app with multiple haproxies
Last modified: 2015-05-14 19:26:07 EDT
Created attachment 786484 [details]
Description of problem:
Create scalable app and send much traffic to it, it will be scaled up automatically, for eg: it has scaled up to 10 gears with 5 haproxy instances running.
Use ab to keep sending request to the app, all the traffics are going to the haproxy instance on the head gear, but no sessions in the secondary haproxy instances(non head gear haproxies). (see attached haproxy_status page for head gear and other gear)
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. Set the multiplier to 2 and max to 5 for haproxy cartridge
2. Create scalable app
3. Use ab to trigger auto-scaling on the scalable app
4. Check the session from the haproxy-status for each haproxy instance.
Only the haproxy on the head gear can receive the requests.
Traffics should go to all the load balancers.
And when disable the haproxy_ctld on the head gear. The app cannot be visited anymore.
Created attachment 786485 [details]
An additional routing layer has to be added that load balances the traffic to the
ha proxies (out of scope for this card). Otherwise, for testing you can have multiple abs sending traffic directly to the different haproxies using their OPENSHIFT_GEAR_DNS at the same time.
According to comment#2
When sending request via ab to the secondary haproxies, the traffics go to the other haproxies.
move bug to verified.