[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Wg-openstack] Minutes from Sept 23rd WG meeting

== Attendees ==

Joao Martins 

Bob Ball

Lars Kurth

Konrad Wilk

Anthony Perard

Michael Glasgow

Jim Fehlig

== Agenda ==

ACTION; Bob & Alexey to arrange meeting (CC Lars and Anthony)

* Bob: Note that Jon remains Nova PTL for next cycle

* Bob: Ant Messerly is currently on sabbatical (back at start of November)

Lars: to send a line to RAX to see whether someone else would pick up Ant's responsibilities

* Bob: Doesn't know what is going on with CI right now - both the Xen Project and XenServer CI has dropped through the floor. Does not know what the root cause is

** See http://zuul.openstack.xenproject.org/scoreboard

** Anthony updated OpenStack CI loop packed to the latest version

Bob: could be that there are lots of new due to OpenStack RC

ACTION: Bob and Anthony to talk and to have a look at failure samples to see whether there are common themes (e.g. such as a new Tempest test) - already looking it. And send updates to the list, if not sure how to fix something.

* Joao's Analysis - respond to the Analysis

** Bob: biggest question is "which" of the missing components need to be tested. At the moment we are pretty memory constrained (8GB is pretty tight).

** Bob: question for the whole group - what is the importance of testing with Heat, Sahara, ...

** Bob: If we enabled Heat, the memory pressure would be larger

** Bob: Are we trying for full Tempest

What do people think? 

** Jim: Goal was to move up to group A - if those items are not needed to reach that goal, then we should not enable them

** Bob: Group A is just a Nova concept - there is no clear info what subset of Tempest needs to be run

** Bob: One could quite comfortably argue that we don''t need Heat tests. Don't see how a driver test would affect Heat for group A

** Joao: What about Neutron

** Bob: Personally I think we have to test Neutron as everyone is expected to use Neutron in future

** Joao: Is Neutron taking up a lot of memory

** Bob: Biggest problem is number of independent Python processes (Neutron has more processes than Nova Network)

** Lars: Issue is that next rung up is 16MB VM's which is 

** Bob: Correct, but have a larger swap disk (would affect performance)

** Bob to Anthony: Why do we have a fixed Dom0 size instead of using ballooning? 

** Joao: Nova wasn't accounting for Dom0 memory - Anthony is not sure whether this is the same issue

** Jim: has seen another report when ballooning is enabled, there may be some issues with ballooning logic. Not sure whether it's libel or libvirt. There was recently another . With fixed memory this issue goes away

* Bob: It is worth revisiting the ballooning question => many of our memory issues would be relaxed if not solved. Anthony does not agree

** Bob: Alternative would be a larger swap disk (will increase run-times slightly)

** Anthony recommends that we should stick with fixed Dom0 size

** Anthony: we use 7.5GB of memory (due to the sheer number of services running in Dom0)

ACTION: talk about memory pressure off-line

For group A the next two priorities would be

** If we had Neutron, then we can argue that we do NOT need support for Heat, Sahara, Ironic, ...

** We do need the live migration tests working

** Michael: what about Cinder? Testing with Cinder already

ACTION: Joao (see below)

** Joao: Waiting for some libvirt changes (peer-2-peer, ...) - Joao will reply with more info in the minutes

** Joao: Libvirt patches are still under "ACTIVE" review by Jim

** Joao: there is a potential OpenStack fix, but the best solution is in libvirt

Had short discussion about OpenStack workflow. Bob suggested to submit a fix for Mitaka release and request back port for Liberty

** Bob: what;s the status of these patches? Is there some help needed from the group?

** Jim/Joao: there have been comments, but additional reviews from Xen community members would help

** Jim: re-basing after virtuoso work is committed would help 

** Joao: Virtuoso work are two different solutions for the same problem. 

** Jim: stat series needs review cycles (should be non-controversial, but slightly large)

** Joao: peer-2-peer issue independent to Virtuoso

ACTION: Lars to ping one of the two Ian's whether they can help out with stars

ACTION: All - Respond to Joao's Analysis in detail

Wg-openstack mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.