| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2018-07-31 | |||
| 12:50:57 | sean-k-mooney | .. need to stop using the server that has only a hdd as my contoler. it takes so long to stack... | |
| 12:51:53 | kashyap | Yeah, SSD for the win! | |
| 12:52:11 | kashyap | sean-k-mooney: BTW, a completely unrelated: Have you had a chance to play with NVMe? | |
| 12:52:58 | sean-k-mooney | that depends. my persoanl laptop and my desktop have nvme drivers | |
| 12:53:16 | sean-k-mooney | none of my server hardware has however | |
| 12:53:48 | sean-k-mooney | i also havenet had a change to play with nvdim or the persitent memory stuff yet. | |
| 12:54:01 | kashyap | Nod; sean-k-mooney, at some point I wonder if we should look at QEMU's VFIO-based block driver for NVMe | |
| 12:54:03 | sean-k-mooney | /drivers/drives/ | |
| 12:54:20 | sean-k-mooney | kashyap: ya that works resonably well. | |
| 12:54:31 | kashyap | The QEMU folks have filed a bug for that; I'm just playing with virtual NVMe driver in QEMU | |
| 12:54:37 | sean-k-mooney | i know you can use it to fake nvme over fabric | |
| 12:54:56 | kashyap | What is 'fabric'? | |
| 12:55:58 | sean-k-mooney | NVMEoF is a way of attaching an nvme driver over the network. such as mapping it over an RDMA over ethernet transport | |
| 12:56:22 | sean-k-mooney | /driver/drive/ i keep making that mistake | |
| 12:56:33 | openstackgerrit | Lenny Verkhovsky proposed openstack/nova master: WIP workaround for RHEL7.5 qemu 2.0 https://review.openstack.org/587409 | |
| 12:56:40 | kashyap | sean-k-mooney: Ah, I see | |
| 13:01:22 | kashyap | sean-k-mooney: Unrelated question about the CirrOS image: | |
| 13:01:49 | kashyap | Despite disabling other data sources, besides "nocloud" | |
| 13:01:50 | kashyap | ><fs> cat /etc/cirros-init/config | |
| 13:01:50 | kashyap | # must set DATASOURCE_LIST | |
| 13:01:50 | kashyap | #DATASOURCE_LIST="nocloud configdrive ec2" | |
| 13:01:50 | kashyap | DATASOURCE_LIST="nocloud" | |
| 13:02:21 | kashyap | When I launch a CirrOS 0.3.5 image, it still does the metadata look up | |
| 13:02:24 | kashyap | Am I missing something? | |
| 13:02:55 | sean-k-mooney | it does not use cloud init as far as i am aware. i think its using glean | |
| 13:03:22 | kashyap | Huh, what is 'glean' now | |
| 13:03:27 | kashyap | And how to disable that? | |
| 13:03:45 | sean-k-mooney | its is an infra replacement for cloud-init because they were missing stuff. | |
| 13:04:11 | sean-k-mooney | are you building the cirros image with diskimage builder locally? | |
| 13:04:45 | kashyap | sean-k-mooney: No-no, it's stock download | |
| 13:05:03 | kashyap | In the past just using: DATASOURCE_LIST="nocloud" reliably worked for me. | |
| 13:05:26 | sean-k-mooney | oh how are setting DATASOURCE_LIST? guestfs? | |
| 13:05:59 | kashyap | Yes, `guestfish --rw -i cirros.qcow2` | |
| 13:06:19 | kashyap | And then `vi /etc/cirros-init/config`, update the config, then run `sync`, and `exit`. | |
| 13:07:06 | sean-k-mooney | right well you would need to check if it is using glean and then disable it if it is | |
| 13:07:54 | kashyap | I couldn't find anything related to 'glean' in /etc; /me looks carefully | |
| 13:09:10 | sean-k-mooney | you should see it in dmesg if its running on boot | |
| 13:09:57 | kashyap | sean-k-mooney: From the shell inside CirrOS: | |
| 13:09:58 | kashyap | $ dmesg | grep -i glean | |
| 13:09:58 | kashyap | $ echo $? | |
| 13:09:58 | kashyap | 1 | |
| 13:10:01 | kashyap | So it's not using it | |
| 13:10:35 | openstackgerrit | Balazs Gibizer proposed openstack/nova master: Improve NeutronFixture and remove unncessary stubbing https://review.openstack.org/587412 | |
| 13:13:09 | gibi | mriedem: This ^^ removes two of your TODOs from the code so you might be interested | |
| 13:21:15 | kashyap | gibi: While Matt looks at that, fancy looking at this: https://review.openstack.org/#/c/567258/ | |
| 13:21:54 | kashyap | It has went through multiple revisions, and all feedback is addressed; and FWIW, has an ACK from sean-k-mooney | |
| 13:22:17 | sean-k-mooney | is overrideing config files with [[post-config|/etc/nova/nova-cpu.conf]] no longer supported by devestack? | |
| 13:23:16 | gibi | kashyap: looking... | |
| 13:25:54 | openstackgerrit | sahid proposed openstack/nova master: hardware: fix memory check usage for small/large pages https://review.openstack.org/532168 | |
| 13:27:33 | kashyap | Thanks! | |
| 13:32:55 | mriedem | gibi: done, comments inline | |
| 13:33:02 | gibi | mriedem: thanks | |
| 13:33:45 | mriedem | sean-k-mooney: it is | |
| 13:36:10 | kashyap | mriedem: Oh, you mean, it is no longer supported? Or it is still supported? | |
| 13:37:40 | sean-k-mooney | mriedem: im trying to get live migration to work but the config overrides where not getting added | |
| 13:37:58 | dansmith | sean-k-mooney: do you need to make those to nova-cpu.conf instead? | |
| 13:38:33 | sean-k-mooney | dansmith: i was making them in nova-cpu.conf nova.conf is not used by the compute agent anymore | |
| 13:38:34 | mriedem | kashyap: post-config is supported | |
| 13:38:41 | mriedem | they need to be paths | |
| 13:39:16 | mriedem | e.g. https://github.com/openstack/nova/blob/master/playbooks/legacy/nova-cells-v1/run.yaml#L38 | |
| 13:39:38 | kashyap | Do you top off your head if this should still work: | |
| 13:39:39 | kashyap | [[post-config|$NOVA_CONF]] | |
| 13:39:39 | kashyap | [DEFAULT] | |
| 13:39:39 | kashyap | force_config_drive = False | |
| 13:39:53 | kashyap | Okay, it should | |
| 13:39:58 | mriedem | https://github.com/openstack/nova/blob/master/playbooks/legacy/nova-caching-scheduler/run.yaml#L37 | |
| 13:39:58 | mriedem | yes | |
| 13:39:59 | sean-k-mooney | dansmith: i made the changes by hand and restarted the agent but im currently trying to figure out a no route issue. | |
| 13:40:24 | kashyap | mriedem: Thanks! | |
| 13:45:12 | openstackgerrit | Matthew Booth proposed openstack/nova master: Ensure attachment cleanup on failure in driver.pre_live_migration https://review.openstack.org/587439 | |
| 13:46:46 | sean-k-mooney | by the way this error message really is not helpful http://paste.openstack.org/show/726939/ | |
| 13:47:51 | sean-k-mooney | it had a route to the host. it could not connect on that port because iptabels was blocking it | |
| 13:59:57 | cdent | in the periodic jobs in nova-compute do we expect the context that it starts with to remain the same for the lifetime of the process? Logs are indicating that it is changing. | |
| 14:00:45 | cdent | http://paste.openstack.org/show/726941/ | |
| 14:02:22 | cdent | rgerganov noticed downstream and we assumed it was some local optimizations, but it appears that maybe the original oslo context is a thread not a green thread | |
| 14:03:14 | cdent | mriedem or dansmith ? ^ | |
| 14:03:37 | dansmith | cdent: that's a known issue | |
| 14:03:42 | dansmith | cdent: it shouldn't change, but it is | |
| 14:03:48 | dansmith | a few people have looked at it | |
| 14:04:03 | cdent | so is it sort of a non-damaging buglet? | |
| 14:04:19 | cdent | is there an existing bug you're aware of? | |
| 14:04:21 | mriedem | it's definitely a serviceability issue | |
| 14:04:22 | mriedem | yes, | |
| 14:04:25 | mriedem | i can find some links | |
| 14:04:27 | dansmith | it's not great because it's a tad confusing, but I'm not sure it's actually a big problem | |
| 14:04:46 | dansmith | if we're ever leaking contexts between request threads then *that* is a huge issue, | |
| 14:04:51 | cdent | I think rado may be sufficiently frustrated by it it want to fix it | |
| 14:04:55 | dansmith | but people would see random failures to look up instances and do things, etc | |
| 14:05:03 | dansmith | and I don't see evidence of that | |
| 14:05:16 | mriedem | gibi has some older changes starting here https://review.openstack.org/#/c/524306/ | |
| 14:05:28 | mriedem | oops that's melwitt's ^ | |
| 14:05:36 | mriedem | https://review.openstack.org/#/c/542891/ is gibi's | |
| 14:06:02 | cdent | thanks mriedem | |
| 14:11:56 | bauzas | mriedem: I don't want to be pedantic about microversions, but look at https://docs.openstack.org/nova/latest/contributor/microversions.html#f1 | |
| 14:12:17 | bauzas | mriedem: I was about +2ing https://review.openstack.org/#/c/583566/ given we merged the master one | |
| 14:12:59 | bauzas | mriedem: but here, a HTTP500 will become a HTTP200 right? | |
| 14:13:16 | bauzas | if so, we should possible change the documentation ^ | |
| 14:13:30 | dansmith | 500->200 is fine right? | |
| 14:13:33 | mriedem | it will change a 500 to a 204 | |
| 14:13:34 | dansmith | that's a bug fix | |