Bug 1375742

Summary: [Docs] [RFE] unable to login to ssui w/ loadbalancer
Product: Red Hat CloudForms Management Engine Reporter: Colin Arnott <carnott>
Component: DocumentationAssignee: Dayle Parker <dayleparker>
Status: CLOSED CURRENTRELEASE QA Contact: Suyog Sainkar <ssainkar>
Severity: urgent Docs Contact:
Priority: urgent    
Version: 5.6.0CC: abellott, adahms, benglish, cpelland, dayleparker, gtanzill, hhudgeon, jhardy, obarenbo, simaishi, ssainkar
Target Milestone: GAKeywords: FutureFeature, ZStream
Target Release: 5.8.0   
Hardware: x86_64   
OS: Linux   
Whiteboard: ssui
Fixed In Version: Doc Type: If docs needed, set a value
Doc Text:
Story Points: ---
Clone Of:
: 1376568 (view as bug list) Environment:
Last Closed: 2017-05-31 01:29:10 UTC Type: Bug
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: CFME Core Target Upstream Version:
Bug Depends On:    
Bug Blocks: 1376568    

Description Colin Arnott 2016-09-13 21:22:46 UTC
Description of problem:
when I login to my UI appliances with my load balancer, sometimes it works fine, but others it signs in and gets immediately kicked out citing a 'session timeout'.

this does not happen when connecting directly to the UI appliance.

this is likely caused by memcached pointing to localhost, in combination with the load balancer.

Version-Release number of selected component (if applicable):
cfme-5.6.1.2

How reproducible:
need loadbalancer

Additional info:
- found during discovery on this bug: 1374398
- when we changed memcached_opts to "-l 0.0.0.0" the evmserverd process did not restart, despite this working when running memcached with systemd locally
 - new memcached_opts likely needed

Comment 2 Joe Rafaniello 2016-09-14 18:07:45 UTC
> this is likely caused by memcached pointing to localhost, in combination with the load balancer.

Hi Colin, are you saying that "-l 0.0.0.0" worked?  I'm not understanding what you think is happening.  Can you provide what you think is the problem with localhost?

From my understanding, if you're hitting apache through the load balancer on the appliance, it will get routed from the apache worker on the appliance to a UI worker on the same appliance which should be able to talk to memcached on localhost.

Comment 3 Colin Arnott 2016-09-14 21:55:15 UTC
I have tried the following configurations:

config:
/session/memcache_server: 127.0.0.1:11211
/session/memcache_server_opts: "-l 127.0.0.1"

results:
default config; everything works


config:
/session/memcache_server: 127.0.0.1:11211
/session/memcache_server_opts: "-l 192.168.0.1 -l 127.0.0.1"

results:
only changed the server_opts; evmserverd starts, but memcached does not listen on the public ip/interface


config:
/session/memcache_server: 192.168.0.1:11211
/session/memcache_server_opts: "-l 192.168.0.1 -l 127.0.0.1"

results:
changed both; evmserverd does not start, citing a Dalii error, because the Dalii gem cannot connect to the memcached service, as it has not bound to the public ip/interface


config:
/session/memcache_server: 192.168.0.1:11211
/session/memcache_server_opts: "-l 192.168.0.61 -l 0.0.0.0 -l 127.0.0.1"

results:
changed both; evmserverd does not start, citing a Dalii error, because the Dalii gem cannot connect to the memcached service, as it has not bound to the public ip/interface

config:
/session/memcache_server: 192.168.0.1:11211
/session/memcache_server_opts: "-l 0.0.0.0 -l 127.0.0.1"

results:
changed both; evmserverd does not start, citing a Dalii error, because the Dalii gem cannot connect to the memcached service, as it has not bound to the public ip/interface


config:
/session/memcache_server: 192.168.0.1:11211
/session/memcache_server_opts: "-l 127.0.0.1"

results:
changed both; evmserverd does not start, citing a Dalii error, because the Dalii gem cannot connect to the memcached service, as it has not bound to the public ip/interface



Sorry for the information spray, but that is what I have tried, and nothing can seem to make memcached bind to the public ip

Comment 4 Colin Arnott 2016-09-14 21:56:47 UTC
Yes, we are hitting apache through a load balancer, so we need everybody to talk to the same memcached on one appliance (we informally picked the database appliance).

Comment 15 Andrew Dahms 2017-03-06 05:07:52 UTC
Moving to 'NEW' while assigned to the default assignee.

Comment 16 Andrew Dahms 2017-04-21 05:21:49 UTC
Assigning to Dayle for review.

Comment 23 Dayle Parker 2017-05-09 07:20:58 UTC
Thanks so much, Suyog!

I've cherry-picked the changes over to master now in https://github.com/ManageIQ/manageiq_docs/pull/348 (also merged).

The change is also live in the 4.2 Deployment Planning Guide, in section 2.5. Using a Load Balancer: https://access.redhat.com/documentation/en-us/red_hat_cloudforms/4.2/html-single/deployment_planning_guide/

Comment 24 Dayle Parker 2017-05-31 01:29:10 UTC
This update has also been published in the CloudForms 4.5 documentation. Closing as CURRENTRELEASE.