Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2019-09-18
16:26:29 sean-k-mooney the config is still the same https://github.com/openstack/nova/blob/master/nova/virt/libvirt/driver.py#L4150-L4181
16:26:59 openstackgerrit melanie witt proposed openstack/nova stable/stein: Add reno about nova-api eventlet monkey-patching and rabbitmq https://review.opendev.org/662095
16:27:16 openstackgerrit Merged openstack/nova stable/stein: neutron: refactor nw info cache refresh out of associate_floating_ip https://review.opendev.org/682181
16:27:23 openstackgerrit Merged openstack/nova stable/stein: Trap and log errors from _update_inst_info_cache_for_disassociated_fip https://review.opendev.org/682182
16:27:32 openstackgerrit Merged openstack/nova stable/stein: Find instance in another cell during floating IP re-association https://review.opendev.org/682183
16:27:39 openstackgerrit Merged openstack/nova stable/stein: Log notifications if assertion in _test_live_migration_force_complete fails https://review.opendev.org/681743
16:28:41 mriedem stephenfin: you mean https://review.opendev.org/#/c/682267/ right?
16:29:03 stephenfin yup
16:29:04 sean-k-mooney mnaser: those feature flags still look wrong to me
16:29:05 mriedem i definitely think we need some functional testing for the libvirt driver that sets cpu extra flags b/c that's 2 regressions when using that config option in the last week or so
16:29:22 mriedem and configuring cpu extra flags seems pretty important for red hat customers...
16:29:24 sean-k-mooney mriedem: it imples we are getting to the extra_flag section or libvirt added them
16:29:25 stephenfin Agreed. I want to flesh out that functional test of mine and use that
16:29:36 stephenfin Just need to figure out where to put it now
16:29:37 sean-k-mooney mriedem: soory that was for mnaser
16:29:46 mnaser sean-k-mooney: yeah it doesn't look right
16:29:54 mnaser but i dont know if virsh dumpxml does some magic
16:29:59 mnaser and doesnt actaully print out what it got from nova
16:30:00 mriedem stephenfin: maybe just a new simple libvirt functoinal test that isn't in the test_numa module or whatever
16:30:01 sean-k-mooney mnaser: was that config form the nova compute log or libvirt via virsh
16:30:04 mnaser no virsh
16:30:10 sean-k-mooney virsh does not
16:30:10 mnaser so i havent blamed nova yet
16:30:17 sean-k-mooney libvirt transforms the xml we pass in
16:30:27 mnaser yeah so this might be transformed
16:30:28 sean-k-mooney e.g. it does not actually use the xml we provide.
16:30:33 mnaser though i wonder why it wouldn't keep host-model there
16:30:33 stephenfin yeah, I'd say so. I just need to check how much of the mocking I need to carry across
16:31:20 mnaser hmm i remember virsh dumpxml used to have a "dumpable" version
16:31:36 sean-k-mooney mnaser: ya i dont know either. this is where i delegate to danpb or kashyap as i read C but dont really want to have to find where that happens
16:31:49 sean-k-mooney mnaser: its not a virsh issue
16:31:59 mnaser ok fair nuff
16:32:12 sean-k-mooney wehn we create teh xml and define the domain libvirt parses it and save an updated version to disk and in memory
16:32:49 sean-k-mooney virsh talks to libvirt over a unix socket and virsh dumpxml give you the rendered version
16:33:11 mnaser oooh
16:33:13 sean-k-mooney libvirt when it ingest the xml fills in a bunch of things like guest pci address that we dont set
16:33:15 mnaser wait so i can do cat on the local disk then
16:33:43 sean-k-mooney there is a copy saved to disk somewhere but you could also just check the nova compute log
16:33:47 sean-k-mooney the xml is saved in it
16:33:55 mnaser yeah not running $world with DEBUG on
16:33:59 mnaser so that'd have to be a little exercise :p
16:34:12 mnaser its only logged when running under debug afaik
16:34:21 sean-k-mooney yes it is
16:34:40 mnaser time to google for the millionth time "how to run an openstack instance on a specific node"
16:34:51 sean-k-mooney anyway i would check the qemu instance log and see what the error is
16:35:00 openstackgerrit Merged openstack/nova-specs master: Update spec: filtering of alloc candidates by forbidden aggregates https://review.opendev.org/675384
16:35:15 sean-k-mooney cat /var/log/libvirt/qemu/instanace...
16:35:33 mnaser yeah but i tihnk its a layer before that because the qemu process doesnt have the flag
16:35:34 mnaser "-cpu IvyBridge-IBRS,ss=on,pcid=on,hypervisor=on,arat=on,tsc_adjust=on,md-clear=on,stibp=on,ssbd=on,xsaveopt=on,pdpe1gb=on"
16:35:54 sean-k-mooney which flag is missing?
16:36:04 sean-k-mooney i though the vm was crashing?
16:36:06 mnaser vmx (and the fact that things are manually added)
16:36:13 mnaser nope, no crashes at all, just literally a flag missing
16:36:23 mnaser virsh capabilities shows vmx
16:36:32 mnaser but booting instances without defining cpu_model does not
16:36:40 sean-k-mooney right but did you turn nested virt on in the host kernel
16:36:42 mnaser does not boot it with vmx then
16:36:46 mnaser yep
16:37:05 sean-k-mooney ok i normally use host-passthough with nested virt so the flag is there
16:37:08 mnaser `virsh capabilities` wouldn't report vmx there if it wasnt on either afaik
16:37:11 mnaser yeah, im trying to avoid that one
16:37:17 sean-k-mooney the model your using might just not included it
16:37:21 sean-k-mooney you could jsut add it
16:37:26 mnaser cat /sys/module/kvm_intel/parameters/nested => Y
16:37:29 sean-k-mooney with the config
16:37:53 mnaser sean-k-mooney: it actually does though -- http://paste.openstack.org/show/777431/
16:37:54 mnaser thats the thing
16:38:09 mnaser i could add it manually but im trying to see if its something we can improve upstream :>
16:38:10 sean-k-mooney the host does
16:38:29 mnaser isnt the output of virsh capabiltiies show the model used by libvirt?
16:38:35 mnaser in this case IvyBridge-IBRS which does include vmx
16:38:36 sean-k-mooney that does not mean the IvyBridge-IBRS has the flag
16:38:43 mnaser wait what
16:38:45 mnaser seriously
16:38:51 sean-k-mooney yes
16:39:07 sean-k-mooney there is an xml file you can check
16:39:10 mnaser i thought virsh capabilities was what would host-model be
16:39:17 sean-k-mooney no
16:39:20 mnaser cpu_map.xml
16:39:28 sean-k-mooney its litrally the host capablites unfiltered
16:39:57 sean-k-mooney the model it reporet is the closet on to your host cpu
16:40:04 mnaser well
16:40:07 mnaser that explains all my confusion
16:40:14 mnaser then i need to add the extra cpu flag
16:40:28 mnaser cpu_map.xml shows nothing actually has vmx flag
16:40:37 mnaser i could _swear_ that it used to be there before
16:40:51 sean-k-mooney where is cpu_map by the way i was looking for it locally
16:41:03 mnaser sean-k-mooney: /usr/share/libvirt/cpu_map.xml
16:41:08 mnaser (on centos at least)
16:41:14 sean-k-mooney ah user share i was looking in /var/lib
16:41:54 mnaser ok im really confused as to how it used to work before but i guess it did
16:42:20 sean-k-mooney it could be a libvirt change
16:42:53 sean-k-mooney or you were using host-passthough and did not no/notice/remember
16:43:20 mnaser no as much as host-passthrough is great im very anti that because live migrations
16:43:51 sean-k-mooney you can use it with live migrations
16:43:59 sean-k-mooney but you need to group things by hardware type
16:44:08 mnaser yeah it gets tricky for new hardware
16:44:09 sean-k-mooney by the way host-model is not relaly better
16:44:40 sean-k-mooney if you use host model and you move for ivybridge to haswell then reboot the vm it wont be abel to migrte back
16:44:51 mnaser yeah im perfectly ok with migrating from old to new but not new to old

Earlier   Later