Bug 1535481 - ODL needs to configure HAProxy so that it recognises isolated controllers
Summary: ODL needs to configure HAProxy so that it recognises isolated controllers
Keywords:
Status: CLOSED DUPLICATE of bug 1488907
Alias: None
Product: Red Hat OpenStack
Classification: Red Hat
Component: puppet-tripleo
Version: 13.0 (Queens)
Hardware: Unspecified
OS: Unspecified
high
high
Target Milestone: beta
: 13.0 (Queens)
Assignee: Tim Rozet
QA Contact: Tomas Jamrisko
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2018-01-17 13:50 UTC by Stephen Kitt
Modified: 2018-10-24 12:36 UTC (History)
9 users (show)

Fixed In Version:
Doc Type: If docs needed, set a value
Doc Text:
Clone Of:
Environment:
N/A
Last Closed: 2018-02-21 15:40:22 UTC
Target Upstream Version:
Embargoed:

Attachments (Terms of Use)
haproxy.cfg (10.32 KB, text/plain)
2018-01-17 15:45 UTC, Tomas Jamrisko 		no flags Details
View All

Description Stephen Kitt 2018-01-17 13:50:51 UTC 
When we end up with a cluster with an isolated leader, HAProxy doesn’t know it shouldn’t forward requests to the isolated node. We should configure the HAProxy probe to react appropriately when the datastore isn’t in a nominal state.
Comment 1 Stephen Kitt 2018-01-17 13:52:15 UTC 
Tomas, can you extract the HAProxy configuration from the setup that gets deployed in our HA CSIT jobs?
Comment 2 Tomas Jamrisko 2018-01-17 15:45:45 UTC 
Created attachment 1382528 [details]
haproxy.cfg

Adding haproxy.cfg
Comment 3 Ofer Blaut 2018-01-18 13:32:41 UTC 
The health check type should try to run a check that  does POST

something like "option httpchk POST ....." so if ODL server is isolated write commands will fail
Comment 4 Stephen Kitt 2018-01-18 14:16:55 UTC 
(In reply to Ofer Blaut from comment #3)
> The health check type should try to run a check that  does POST
> 
> something like "option httpchk POST ....." so if ODL server is isolated
> write commands will fail

That would be a good first step, but POSTs to a failed ODL server take a *long* time to time out so it would still leave us in a bad state for quite a while (unless we tell HAP to expect a response rapidly and fail the node otherwise).
Comment 8 Tim Rozet 2018-02-21 15:40:22 UTC 

*** This bug has been marked as a duplicate of bug 1488907 ***
Note You need to log in before you can comment on or make changes to this bug.