openstack-manuals

Bug #1683585
Comment #11

Comment 11 for bug 1683585

Revision history for this message

Matt Riedemann (mriedem) wrote on 2017-04-24:

#11

Here is the IRC conversation we had in #openstack-nova today:

(2:21:23 PM) mriedem: coreycb: can you comment on the state of the nova-placement-api package wrt https://bugs.launchpad.net/openstack-manuals/+bug/1683585 /
(2:21:24 PM) mriedem: ?
(2:21:25 PM) openstack: Launchpad bug 1683585 in openstack-manuals "Placement endpoint configuration incorrect as written" [High,Confirmed] - Assigned to Brian Moss (bmoss)
coreycb coreywright
(2:21:54 PM) coreycb: mriedem, sure, taking a look
(2:22:15 PM) mriedem: smcginnis: sounds like you should be able to do the workaround in that bug report by deleting your endpoints and recreating them with the 8778 port
(2:22:33 PM) smcginnis: mriedem: Just finished recreating them, let's see how this goes.
(2:22:41 PM) ***mriedem crosses fingers
(2:24:22 PM) coreycb: smcginnis, i've always created them with the port numbers, i need to what chris was talking about
(2:24:35 PM) coreycb: that is, i need to look into
(2:25:20 PM) cdent: coreycb: I think the fundamental issue was that I said (in a review of the install guide): in the long term we want to use prefixes like /placement instead of ports
(2:25:26 PM) cdent: but that got translated into changing the docs
(2:25:36 PM) smcginnis: coreycb: Great. Yeah, I was just following the docs.o.o ubuntu instructions, and that doesn't work right now.
(2:25:37 PM) cdent: causing a mismatch with the packges
(2:26:02 PM) smcginnis: Hmm, not I see this in the scheduler log: The placement API endpoint not found. Placement is optional in Newton, but required in Ocata. Please enable the placement service before upgrading.
(2:26:37 PM) mriedem: smcginnis: you're getting an endpoint not found error from ksa now
(2:26:59 PM) mriedem: smcginnis: the service type and endpoint name are 'placement' right?
(2:27:11 PM) coreycb: cdent, yep
(2:27:55 PM) mriedem: cdent: unless the ubuntu packages are going to change to take the port number out of the apache config it seems we should get the install guide updated to create the endpoint with the port number
(2:28:07 PM) cdent: mriedem: yes
(2:28:15 PM) coreycb: mriedem, cdent: yes agreed
(2:28:25 PM) cdent: that's pretty much where asettle and I landed when talking about it last week
(2:28:30 PM) mriedem: if canonical changed the apache config in a new package release, that would be confusing (at least for ocata)
(2:28:50 PM) mriedem: you could probably mirate the config somehow at install/update time for the new package, but...
(2:28:54 PM) mriedem: *migrate
(2:29:01 PM) smcginnis: Well, if our long term goal is to get rid of ports, shouldn't we make sure new services aren't getting set up that way by default?
(2:29:36 PM) mriedem: smcginnis: sure, but then canonical has to change the config in their package and provide a migration i'd think
(2:29:43 PM) mriedem: which is probably a PITA after ocata GA'ed
(2:29:58 PM) coreycb: mriedem, smcginnis: yeah we'd definitely want to do that on a release boundary if we did

Here is the IRC conversation we had in #openstack-nova today:

(2:21:23 PM) mriedem: coreycb: can you comment on the state of the nova-placement-api package wrt https://bugs.launchpad.net/openstack-manuals/+bug/1683585 /
(2:21:24 PM) mriedem: ?
(2:21:25 PM) openstack: Launchpad bug 1683585 in openstack-manuals "Placement endpoint configuration incorrect as written" [High,Confirmed] - Assigned to Brian Moss (bmoss)
coreycb coreywright 
(2:21:54 PM) coreycb: mriedem, sure, taking a look
(2:22:15 PM) mriedem: smcginnis: sounds like you should be able to do the workaround in that bug report by deleting your endpoints and recreating them with the 8778 port
(2:22:33 PM) smcginnis: mriedem: Just finished recreating them, let's see how this goes.
(2:22:41 PM) ***mriedem crosses fingers
(2:24:22 PM) coreycb: smcginnis, i've always created them with the port numbers, i need to what chris was talking about
(2:24:35 PM) coreycb: that is, i need to look into
(2:25:20 PM) cdent: coreycb: I think the fundamental issue was that I said (in a review of the install guide): in the long term we want to use prefixes like /placement instead of ports
(2:25:26 PM) cdent: but that got translated into changing the docs
(2:25:36 PM) smcginnis: coreycb: Great. Yeah, I was just following the docs.o.o ubuntu instructions, and that doesn't work right now.
(2:25:37 PM) cdent: causing a mismatch with the packges
(2:26:02 PM) smcginnis: Hmm, not I see this in the scheduler log: The placement API endpoint not found. Placement is optional in Newton, but required in Ocata. Please enable the placement service before upgrading.
(2:26:37 PM) mriedem: smcginnis: you're getting an endpoint not found error from ksa now
(2:26:59 PM) mriedem: smcginnis: the service type and endpoint name are 'placement' right?
(2:27:11 PM) coreycb: cdent, yep
(2:27:55 PM) mriedem: cdent: unless the ubuntu packages are going to change to take the port number out of the apache config it seems we should get the install guide updated to create the endpoint with the port number
(2:28:07 PM) cdent: mriedem: yes
(2:28:15 PM) coreycb: mriedem, cdent: yes agreed
(2:28:25 PM) cdent: that's pretty much where asettle and I landed when talking about it last week
(2:28:30 PM) mriedem: if canonical changed the apache config in a new package release, that would be confusing (at least for ocata)
(2:28:50 PM) mriedem: you could probably mirate the config somehow at install/update time for the new package, but...
(2:28:54 PM) mriedem: *migrate
(2:29:01 PM) smcginnis: Well, if our long term goal is to get rid of ports, shouldn't we make sure new services aren't getting set up that way by default?
(2:29:36 PM) mriedem: smcginnis: sure, but then canonical has to change the config in their package and provide a migration i'd think
(2:29:43 PM) mriedem: which is probably a PITA after ocata GA'ed
(2:29:58 PM) coreycb: mriedem, smcginnis: yeah we'd definitely want to do that on a release boundary if we did