Tuesday, 2019-03-12

« Monday, 2019-03-11 Index Wednesday, 2019-03-13 »
openstackgerritAlexey Khodos proposed openstack/cinder master: Refactored NexentaStor5 driver  https://review.openstack.org/58623900:18
*** tejdeep has joined #openstack-cinder00:41
*** erlon has joined #openstack-cinder00:41
*** tejdeep_ has joined #openstack-cinder00:43
*** tejdeep has quit IRC00:43
*** tejdeep has joined #openstack-cinder00:44
*** dave-mccowan has joined #openstack-cinder00:54
*** enriquetaso has quit IRC00:55
*** erlon has quit IRC00:57
*** tejdeep has quit IRC00:57
*** whoami-rajat has joined #openstack-cinder01:12
*** itlinux has joined #openstack-cinder01:25
openstackgerritRyan Liang proposed openstack/cinder stable/queens: VNX: update sg in cache  https://review.openstack.org/64261301:31
*** mvkr has quit IRC01:36
*** mvkr has joined #openstack-cinder01:38
*** gnufied has quit IRC01:41
*** erlon has joined #openstack-cinder01:48
*** mvkr has quit IRC01:49
*** itlinux has quit IRC01:51
*** mvkr has joined #openstack-cinder01:52
openstackgerritMerged openstack/cinder master: Adds revert to snapshot feature to support matrix  https://review.openstack.org/62797102:20
*** openstackstatus has quit IRC02:22
*** openstack has joined #openstack-cinder02:26
*** ChanServ sets mode: +o openstack02:26
*** Dinesh_Bhor has quit IRC02:49
*** lixiaoy1 has joined #openstack-cinder02:58
*** lixiaoy1 has quit IRC03:03
*** erlon has quit IRC03:35
*** dave-mccowan has quit IRC03:40
*** udesale has joined #openstack-cinder03:49
*** mchlumsky has quit IRC03:58
*** rosmaita has left #openstack-cinder04:11
*** udesale has quit IRC04:45
*** udesale has joined #openstack-cinder04:46
*** abhishekk has joined #openstack-cinder04:57
*** abhishekk has quit IRC04:57
*** markvoelker has joined #openstack-cinder05:06
*** avishay has joined #openstack-cinder05:17
avishayHi all, it looks like I have introduced some regression with the WWN verification patch in os-brick05:18
avishayhttp://openstack-logs.purestorage.com/09/636709/6/check/PureISCSIDriver-tempest-dsvm-xenial-aio-multipath-chap/414e0f8/logs/screen-c-vol.txt.gz05:18
avishayFor some reason get_sysfs_wwn doesn't always return a WWN, and therefore the verification fails05:19
avishayPossible solutions are: (1) skip the validation if no WWN is found, (2) use get_scsi_wwn() instead of get_sysfs_wwn(), (3) revert the WWN verification patch in os-brick until Train opens05:21
avishaypatrickeast, This only affects Pure Storage because it's the only driver that returns the expected WWN on initialize_connection. We can revert that patch if desired as an alternate short-term fix.05:22
*** sapd1 has joined #openstack-cinder05:24
avishayI would say that the safest is probably to revert the WWN verification in os-brick05:24
whoami-rajatavishay: 3rd option seems best for now.05:24
avishaywhoami-rajat: Yes. What's the process for reverting? Open a bug and submit a revert patch?05:25
whoami-rajatavishay: you can directly push the revert patch. but since os-brick is released[1], i'm not sure how we can add the revert to it. maybe need a new release. but atleast for now you can push the patch for it.05:30
whoami-rajat[1] https://review.openstack.org/#/c/641863/05:30
avishaywhoami-rajat: OK.  I will submit the revert patch for now and I guess we'll have to wait until people wake up?05:36
openstackgerritAvishay Traeger proposed openstack/os-brick master: Revert "Verify WWN of connected iSCSI devices if passed"  https://review.openstack.org/64264805:38
avishaywhoami-rajat: ^^^05:39
whoami-rajatavishay: yeah, i think smcginnis  would provide better inputs here.05:46
avishaywhoami-rajat: OK thanks a lot05:50
whoami-rajatavishay: np!05:56
*** gary_perkins has quit IRC06:08
*** irclogbot_0 has quit IRC06:09
*** andreaf has quit IRC06:09
*** gary_perkins has joined #openstack-cinder06:10
*** irclogbot_0 has joined #openstack-cinder06:10
*** andreaf has joined #openstack-cinder06:12
*** e0ne has joined #openstack-cinder06:18
openstackgerritRajat Dhasmana proposed openstack/cinder master: Automate generation of snapshots api-ref samples  https://review.openstack.org/64265806:49
*** Luzi has joined #openstack-cinder06:53
*** dpawlik has joined #openstack-cinder07:30
*** pcaruana has joined #openstack-cinder07:39
*** pcaruana has quit IRC07:43
*** pcaruana has joined #openstack-cinder07:43
*** tkajinam has quit IRC08:07
*** e0ne has quit IRC08:17
*** e0ne has joined #openstack-cinder08:23
*** e0ne has quit IRC08:39
*** sapd1 has quit IRC08:47
*** sahid has joined #openstack-cinder08:48
openstackgerritYury Kulazhenkov proposed openstack/cinder master: Add support for VxFlex OS 3.0 to VxFlex OS driver  https://review.openstack.org/63996309:29
*** mpasserini has joined #openstack-cinder09:45
mpasseriniHi, I just posted a comment to this patch, I hope it's the right place: https://review.openstack.org/#/c/595372/209:46
mpasseriniThis causes problems when deleting backups created before updating configuration, the error shown is like: "Delete backup aborted, the backup service currently configured [cinder.backup.drivers.swift] is not the backup service that was used to create this backup [cinder.backup.drivers.swift.SwiftBackupDriver]." It requires manual fixing in the DB like this: "update backups set service='cinder.backup.drivers.swift' where se09:46
mpasserini='cinder.backup.drivers.swift.SwiftBackupDriver';"09:46
mpasseriniIs there a better place where to track the issue? Bugzilla?09:46
whoami-rajatmpasserini: Hi, thanks for the insights, you can report a bug on launchpad [1] (right side)10:02
whoami-rajat[1] https://launchpad.net/cinder10:02
openstackgerritAvishay Traeger proposed openstack/os-brick master: Revert "Verify WWN of connected iSCSI devices if passed"  https://review.openstack.org/64264810:04
mpasseriniwhoami-rajat: ok thanks!10:05
*** e0ne has joined #openstack-cinder10:16
*** dviroel_ has joined #openstack-cinder10:32
*** lpetrut has joined #openstack-cinder10:33
*** luizbag has joined #openstack-cinder10:44
*** bal has joined #openstack-cinder10:59
*** carlos_silva has joined #openstack-cinder11:11
*** sahid has quit IRC11:17
*** sahid has joined #openstack-cinder11:20
*** thgcorrea has joined #openstack-cinder11:20
*** dave-mccowan has joined #openstack-cinder11:23
*** abishop has joined #openstack-cinder11:53
*** rosmaita has joined #openstack-cinder12:14
*** markvoelker has quit IRC12:18
*** e0ne has quit IRC12:36
*** avishay has quit IRC12:37
*** udesale has quit IRC12:50
*** udesale has joined #openstack-cinder12:51
*** mriedem has joined #openstack-cinder12:53
*** enriquetaso has joined #openstack-cinder13:02
*** FlorianFa has quit IRC13:03
*** sapd1 has joined #openstack-cinder13:03
smcginnismpasserini: What version are you running. That issue was fixed (or at least was supposed to have been).13:07
mpasserinismcginnis I'm at queens13:09
mpasseriniopenstack-cinder-12.0.4-3.el7ost.noarch13:09
smcginnisHmm, that should have it.13:10
smcginnishttps://github.com/openstack/cinder/commit/7bae5759c872d4536ee29887eb6567f668e75eec#diff-c821b31d1ba94112c25a0bd41d55d5fb13:10
smcginnisOh!13:11
smcginnismpasserini: I finally read it close enough. Your service is misconfigured.13:12
*** anks2k has joined #openstack-cinder13:16
mpasserinismcginnis: the system was working fine with the driver spelled as cinder.backup.drivers.swift.SwiftBackupDriver until we did the ugprade to Queens. With Queens the spelling was corrected in the code to cinder.backup.drivers.swift, so our backups created with the old method cannot be deleted anymore13:16
mpasseriniwe had to change the 'service' column in the cinder backup DB for those old backups in order to make their deletion work13:17
*** lseki has joined #openstack-cinder13:21
smcginnismpasserini: That's actually backwards. The format prior to Queens allowed using just the module name (cinder.backup.drivers.swift) but starting with Queens it got rid of that backwards compatibility so you actually have to provide the class name for the backup driver (cinder.backup.drivers.swift.SwiftBackupDriver).13:21
*** dave-mccowan has quit IRC13:22
mpasserinioh I see, so I should change this? [root@controller-3 ~]# grep drivers.swift /etc/cinder/cinder.conf13:26
mpasserinibackup_driver = cinder.backup.drivers.swift13:26
smcginnismpasserini: Correct13:26
smcginnisYou will also need to restart the cinder-backup service after editing /etc/cinder.conf13:27
smcginnisAnd it sounds like another DB update to revert back the one you had done.13:27
*** e0ne has joined #openstack-cinder13:28
mpasseriniI see ,then it's a problem with the puppet module version delivered with RHOSP13, which I'm using13:28
mpasserini[root@controller-3 ~]# grep cinder.backup.drivers.swift /etc/puppet/modules/cinder/manifests/backup/swift.pp13:29
mpasserini  $backup_driver                = 'cinder.backup.drivers.swift',13:29
smcginnismpasserini: Hmm, yeah. They should not be setting it to that.13:29
mpasseriniThe last puppet module includes the right driver https://github.com/openstack/puppet-cinder/blob/4aaac7a5b42e4aa969ddef2205d9a4aa1d2db84c/manifests/backup/swift.pp#L7813:32
mpasserinithanks!13:32
smcginnismpasserini: No problem. Glad it got figured out!13:33
*** dave-mccowan has joined #openstack-cinder13:35
*** avishay has joined #openstack-cinder13:39
avishaysmcginnis: Hey13:40
*** enriquetaso has quit IRC13:48
*** lennyb has quit IRC13:51
*** openstack has joined #openstack-cinder15:42
*** ChanServ sets mode: +o openstack15:42
avishayThanks guys and sorry for that15:44
jungleboyjavishay:  It is ok.  Better safe than sorry.15:44
avishayAlso better late than never :P15:45
*** e0ne has quit IRC15:45
*** enriquetaso has quit IRC15:46
smcginnisjungleboyj: Might be good to get it going right away.15:47
*** enriquetaso has joined #openstack-cinder15:47
jungleboyjsmcginnis:  Ok.  Cool.  I thought that made the most sense but wasn't sure with the way you worded things.  I will keep pushing this forward.15:47
jungleboyjwhoami-rajat:  Has started the merge so I will request the release as soon as it merges.15:49
whoami-rajatjungleboyj: after your +2, thought its better to merge it fast. :)15:50
jungleboyjwhoami-rajat:  Yeah, makes sense.15:51
hemna_mep15:53
*** sapd1 has quit IRC15:53
jungleboyjI just hadn't gotten there yet in my processing of the situation.15:54
*** avishay has quit IRC15:54
*** e0ne has joined #openstack-cinder15:58
*** sahid has quit IRC16:00
whoami-rajate0ne:  since we merged the ScaleIO renaming change in os-brick[1] , and it's blocked in cinder[2], won't it cause any issue?16:01
whoami-rajat[1] https://review.openstack.org/#/c/635530/16:01
whoami-rajat[2] https://review.openstack.org/#/c/634397/16:01
*** sapd1 has joined #openstack-cinder16:01
*** sahid has joined #openstack-cinder16:02
e0newhoami-rajat: we reached requirements freeze for Stein, so it should be safe16:02
whoami-rajatok, i think the os-brick change is backward compatible so it seems safe.16:03
whoami-rajate0ne: ok thanks.16:03
*** enriquetaso has quit IRC16:05
jungleboyje0ne:  Though we are going to request a FFE for requirements for os-brick for this change:16:05
jungleboyjhttps://review.openstack.org/#/c/642648/216:05
jungleboyjOh, but that merged before the previous os-brick release.  So, if it is backward compatible then we should be ok.16:06
*** KeithMnemonic has quit IRC16:07
*** anks2k has left #openstack-cinder16:07
*** KeithMnemonic has joined #openstack-cinder16:08
*** sapd1 has quit IRC16:08
whoami-rajatjungleboyj: some part of backward compatibility exists in cinder patch also[1] (as smcginnis suggested)16:11
whoami-rajat[1] https://review.openstack.org/#/c/634397/6/cinder/volume/manager.py@18516:11
jungleboyjRight, but the old names will work in os-brick without the Server side merging.16:13
hemna_why do we have an os-brick specific patch for a cinder driver?16:14
hemna_review url ?16:15
hemna_https://review.openstack.org/#/c/640875/16:17
hemna_that one?16:17
hemna_why is there branding specific names in the connection properties at all?16:18
whoami-rajathemna_: it's for renaming the connector https://review.openstack.org/#/c/635530/16:18
hemna_no customer EVER sees that?16:18
hemna_why not make it just generic16:18
hemna_it's odd to me that there is branding as part of a connector16:19
jungleboyjhemna_: Good point.16:19
hemna_smh16:19
hemna_now we have special cases in os-brick because of branding....16:20
* hemna_ has a sad16:20
eharneywhy is the class named VxFlexOsConnector when all of the text calls it "VxFlexOS"?  uhy16:20
* eharney just reviewed https://review.openstack.org/#/c/634397/ and kinda thinks we should just name it "EMC Thing #4" and give up on the marketing name changes16:21
hemna_yah I agree16:21
hemna_they even changed the name of the protocol in the connection_properties that gets shipped between cinder and nova16:21
hemna_smh16:21
hemna_because....branding16:21
whoami-rajathemna_:  i think due to custom driver protocols.16:22
jungleboyjI should have pushed back against that.16:22
eharneythey could have just left the protocol as "scaleio".16:22
hemna_eharney: +116:22
hemna_well, I guess I should blame myself for not doing more reviews16:23
hemna_but I would have -1'd those for sure16:23
* jungleboyj was doing too many reviews to put enough thought into that.16:23
hemna_I get renaming the driver in cinder16:24
hemna_we did that at HP -> HPE a few years back16:24
hemna_as it's in the public documentation eventually16:24
hemna_I kinda think we should revert that os-brick change for the protocol16:25
* hemna_ wonders if the connector mapping can ever get removed in the future16:26
hemna_even if it is deprecated16:26
hemna_re: detaching a volume that's been attached for a long time16:27
hemna_if we remove that deprecated mapping, I think it will fail to detach16:27
hemna_as os-brick won't be able to find the right connector16:27
hemna_:(16:27
whoami-rajatjungleboyj: wasn't the disco driver removed from cinder?16:27
hemna_whoami-rajat: can we just make the rebranded cinder driver use the same 'scaleio' protocol16:28
hemna_that will clean up os-brick a bit16:28
jungleboyjwhoami-rajat: Yes. It was.16:28
hemna_and just add a comment in the cinder driver that you have to keep the protocol name the same for backwards and forwards compatibility16:29
whoami-rajathemna_: yes, was looking for a case like that. they've also changed all the connection properties https://review.openstack.org/#/c/635530/6/os_brick/initiator/connectors/vxflexos.py16:29
hemna_ugh16:29
hemna_that's unfortunate16:30
hemna_:(16:30
whoami-rajatjungleboyj:  so i think this code isn't used anywhere ? https://github.com/openstack/os-brick/blob/master/os_brick/initiator/connectors/disco.py16:30
hemna_I think that once you have a protocol mapping in os-brick, it can't be removed16:30
jungleboyjThis sounds like a mess we should discuss in Denver.16:31
hemna_I think we should fix this prior to shipping the next os-brick (if we can)16:32
hemna_https://github.com/openstack/os-brick/blob/stable/rocky/os_brick/initiator/connector.py#L10416:32
hemna_the only real change needed in the connector.py would be to change that line16:33
hemna_to the new connector name16:33
hemna_once the current state ships, we can't remove that mapping, as the new protocol name won't work16:34
jungleboyjI am open to doing whatever is best for os-brick going forward.16:36
*** udesale has quit IRC16:37
hemna_it's going to require making changes in the rebranded cinder driver as well16:37
hemna_https://review.openstack.org/#/c/640875/4/os_brick/initiator/connectors/vxflexos.py16:38
hemna_:(16:39
whoami-rajathemna_: the driver isn't renamed in cinder, the nova renaming required os-brick release (which is done) but still nova patch is on hold[1], if we just revert the os-brick change, what will be the issue?16:40
whoami-rajat[1] https://review.openstack.org/#/c/634866/316:40
hemna_yah that should be -2'd16:41
hemna_what a mess16:41
jungleboyjwhoami-rajat:  That would be nice.16:41
jungleboyjLooks like Eric asked the right question.16:41
hemna_yah I think we should revert these changes16:44
hemna_do we have contact with Yury?16:45
whoami-rajathemna_: not sure, can't find him now,  the revert can be proposed by anybody i guess?16:46
hemna_yah16:46
jungleboyjWe should.  We can get hold of walshh also.16:46
hemna_either revert or new patches to fix the protocol16:46
jungleboyjShe, of course, is not on IRC.16:46
whoami-rajatjungleboyj: i'm still waiting for walshh's answer to my previous query :)16:48
whoami-rajathemna_: i will try contacting yury tomorrow again. else we can continue with either of the approach.16:50
jungleboyjI would vote for a revert, get this mess out of here since the merge has been blocked on the Nova side.16:50
whoami-rajatjungleboyj: ++16:50
jungleboyjAlso since we haven't merged the server side and then work things out properly in Train.16:51
hemna_ok that's probably the best at this point16:53
*** e0ne has quit IRC16:55
jungleboyjwhoami-rajat:  You are going to follow up with Yuri tomorrow and explain why we are doing this16:57
*** enriquetaso has joined #openstack-cinder16:58
whoami-rajatjungleboyj: sure... (if i find him in IST tz).16:59
jungleboyjOk.  Who is going to propose the reversion patch?16:59
jungleboyjhemna_:   ?16:59
hemna_sure I can work on that17:00
jungleboyjCool.  I will help to approve and then once everything has merged I will propose the new os-brick release.17:00
whoami-rajatjungleboyj: also i found this [1] . RSD driver is on hold in cinder but respective os-brick changes are merged. (this probably looks more safe as more than the vxflex one)17:01
whoami-rajat[1] https://review.openstack.org/#/c/620250/17:01
whoami-rajatjungleboyj: can i also add this as meeting agenda for tomorrow ?17:01
whoami-rajat^ the cinder - osbrick gaps part discussion17:02
jungleboyjwhoami-rajat:  Yeah, that is a good topic.17:02
jungleboyjI think the RSD connector is fine as it is an addition that was needed to support their driver.  The code just isn't used now.17:03
jungleboyjUnless hemna_  looks at that and disagrees.17:03
* jungleboyj is glad to have our os-brick expert back to some extent.17:03
whoami-rajatjungleboyj:  thanks. added the topic.17:04
*** enriquetaso has quit IRC17:05
*** enriquetaso has joined #openstack-cinder17:05
*** eandersson_ has joined #openstack-cinder17:07
jungleboyjwhoami-rajat:  Thanks.17:17
jungleboyjI have alerted prometheanfire that an FFE will be coming for requirements.17:18
thgcorreaeharney, I changed the commit message to better match my fix on https://review.openstack.org/#/c/633596/ that you commented with erlon17:27
eharneythgcorrea: thanks, will take a look17:28
*** mpasserini has left #openstack-cinder17:39
openstackgerritMerged openstack/os-brick master: Revert "Verify WWN of connected iSCSI devices if passed"  https://review.openstack.org/64264817:39
hemna_back17:51
hemna_so that one I think is fine, as it's a brand new connector for nvme17:53
hemna_new connector, new protocol17:53
hemna_hrmm17:53
hemna_new protocol, but same connector.....17:54
hemna_"* Add a new protocol "NVMEOF" mapping to allow both Cinder and Nova to17:54
hemna_  identify and use the nvme connector.17:54
hemna_"17:54
hemna_why17:54
hemna_there is already an nvme protocol17:55
*** zhubx has quit IRC17:55
*** zhubx has joined #openstack-cinder17:55
hemna_unless I'm missing something, I think that one was also unnecessary17:55
hemna_whoami-rajat: is there an associated nova patch to accomodate nvmeof protocol as well?17:56
*** henriqueof has joined #openstack-cinder17:57
whoami-rajathemna_: don't think so17:58
hemna_smh17:58
hemna_wtf17:58
whoami-rajathemna_: where do we use the nvme connector ? Spdk driver ?17:59
hemna_I'm trying to find it17:59
jungleboyjwhoami-rajat:  It will be in the RSD driver when they get that in place.17:59
jungleboyjhemna_:  So you are saying because the NVME-of driver just redirects to the NVME driver there was no need to make a new connector?18:00
hemna_spdk.py looks like18:00
hemna_no18:00
hemna_there is no need to create the new protocol18:00
hemna_https://review.openstack.org/#/c/620250/10/os_brick/initiator/__init__.py18:00
whoami-rajatjungleboyj: they added nvme-of for rsd, nvme existed previously18:00
jungleboyjHmmm, didn't realize spdk was using it too.18:01
hemna_https://review.openstack.org/#/c/620250/10/os_brick/initiator/connector.py18:01
hemna_both NVME and NVMEOF point to the same connector18:01
jungleboyjAh, I see.  Should have said protocol.18:01
jungleboyjBut I am following.18:01
*** e0ne has joined #openstack-cinder18:02
whoami-rajatI think they added the new connector due to the additional options used by rsd18:03
hemna_well that patch doesn't have a new connector18:03
hemna_they just updated the same connector18:03
hemna_is there a patch in cinder to use the NVMEOF protocol?18:04
hemna_because I don't see it in the current code for cinder18:04
hemna_making that os-brick patch useless18:04
eharneyhemna_: https://review.openstack.org/#/c/621465/18:05
hemna_eharney: ah yah, just found that too18:05
whoami-rajathemna_: oh yes, maybe they wanted to create a new connector and inherit the nvme connector but the end patch merged seems wrong.18:05
hemna_yah this is borked18:06
hemna_as well18:06
hemna_:(!!18:06
*** e0ne has quit IRC18:06
hemna_there is nothing special in the nvme.py connector that even looks at the nvmeof protocol18:07
hemna_so, yah, that shouldn't have happened either18:07
* jungleboyj is sad18:09
hemna_what a mess18:10
whoami-rajathemna_: shouldn't this change in nvme connector have affected the spdk driver ?18:10
hemna_so18:11
hemna_the nvme.py connector changes adds a new 'system uuid' in the connector dict that gets sent back to cinder18:12
hemna_and the nvme.py connector also looks for host_nqn18:12
hemna_if it's there, it uses it18:12
hemna_so I'm guessing that the spdk driver in cinder should still work18:13
hemna_host_nqn is returned by the new rsd.py cinder driver, but not spdk.py driver18:15
*** erlon has quit IRC18:15
*** sahid has quit IRC18:17
*** e0ne has joined #openstack-cinder18:18
whoami-rajathemna_: only if it's found in the connection_properties, this seems to be written to handle both drivers18:18
hemna_yah, there is no need for the nvmeof protocol18:18
* hemna_ wants to cry18:18
hemna_this is such a mess now18:18
whoami-rajathemna_: :(18:19
jungleboyj:-(  Is this something we can fix before Stein goes out?18:19
hemna_what's the timing ?18:20
jungleboyjWell, it depends how long they will take requirements FFEs.18:21
*** e0ne_ has joined #openstack-cinder18:21
hemna_so I think I have several patches to revert just for the scaleio issue18:21
hemna_https://github.com/openstack/os-brick/commit/6b60614afaf88819ce92a2d0ed269035a3c8b26118:21
hemna_https://github.com/openstack/os-brick/commit/04c6c0cc291bd8c472b2a24a333085739eb9f87718:21
hemna_those 218:21
whoami-rajathemna_: how about removing the nvmeof from connector names and making changes in the rsd driver to use nvme?18:21
*** e0ne has quit IRC18:22
hemna_and I think we can get away with simply removing the mapping in os-brick for nvmeof18:22
hemna_like whoami-rajat just mentioned, change the rsd driver to use nvme protocol18:22
hemna_I already -1'd that cinder rsd.py patch18:23
hemna_but that's a new driver, so it shouldn't land now anyway18:23
hemna_did the scaleio driver rebranding land in cinder?18:23
jungleboyjhemna_:  No.18:23
hemna_do you have that review url handy?18:24
jungleboyjhemna_:  In another meeting.  Hold on.18:25
hemna_ok18:25
whoami-rajathemna_: although I prefer nvmeof name over nvme as it makes more sense here, but I guess that's a topic for other day!18:25
hemna_I just want to get my ducks in a row, before I go about reverting commits18:25
hemna_and get a plan in place18:25
hemna_whoami-rajat: yah18:25
*** gmann is now known as gmann_afk18:26
whoami-rajathttps://review.openstack.org/#/q/owner:yury.kulazhenkov%2540dell.com+status:open18:27
whoami-rajathemna_: ^18:27
hemna_wtf18:32
hemna_https://review.openstack.org/#/c/634397/6/cinder/volume/drivers/dell_emc/vxflexos/driver.py18:32
hemna_that patch still uses the scaleio connector protocol18:32
hemna_line 92818:32
*** angela-s has joined #openstack-cinder18:32
jungleboyjwhoami-rajat:  Thanks for finding the link.18:33
angela-ssmcginnis Hi Sean, can you have a look at the Newton backport? Jay has already approved. Thank you! https://review.openstack.org/#/c/640624/18:33
whoami-rajatjungleboyj: np!18:34
hemna_ok, so the cinder driver still uses 'scaleio' as the protocol18:34
hemna_so we just need to backout the os-brick commits18:35
hemna_and then they can accomodate for a new patch to use the scaleio protocol18:35
hemna_and rid os-brick of the mapping18:35
hemna_yah the VXFLEXOS protocol isn't even used in his rebranding patches in cinder18:37
hemna_smh18:37
hemna_and none of the cinder and nova changes have landed18:37
hemna_so it's safe to revert IMHO18:37
openstackgerritWalter A. Boring IV (hemna) proposed openstack/os-brick master: Revert "Fix VxFlexOs KeyError after upgrade"  https://review.openstack.org/64284918:38
whoami-rajathemna_: is it worth considering to have names like emc connector for better generalized connectors? Also for the filename atleast18:39
*** pcaruana has quit IRC18:39
hemna_maybe, but we have existing protocols that I don't want to have to create a mapping18:39
hemna_which is part of the problem here18:39
hemna_creates a mess of magic18:40
*** e0ne_ has quit IRC18:41
openstackgerritWalter A. Boring IV (hemna) proposed openstack/os-brick master: Revert "rename ScaleIO connector to VxFlex OS"  https://review.openstack.org/64285218:41
hemna_ok both of the revert patches are up in os-brick18:42
whoami-rajathemna_: ok. Thanks for looking into all these.18:42
hemna_np18:43
hemna_glad we caught it before the release went out18:43
whoami-rajathemna_: yeah, maybe merging os brick changes before driver merge isn't a good idea.18:44
hemna_it's a little bit of a chicken and egg issue really18:45
jungleboyjhemna_:  Sorry for helping to make the mess.18:45
jungleboyjhemna_: whoami-rajat We should use cross project dependencies to help control that.18:45
whoami-rajatjungleboyj: ++18:47
hemna_so https://review.openstack.org/#/c/620250/18:48
hemna_I don't think we should revert the whole patch18:49
hemna_but add a new patch that just removes NVMEOF protocol18:49
hemna_https://review.openstack.org/#/c/620250/10/os_brick/initiator/connector.py18:49
hemna_basically undoing that18:49
hemna_and force the rsd driver to use nvme18:49
whoami-rajathemna_: agreed18:50
eharneynvme isn't the same thing as nvmeof..?18:51
hemna_the technology?  got me18:51
hemna_both of the connector protocols point to the same exact os-brick connector18:51
hemna_it might make sense if there was something specific about the nvmeof protocol that required an nvmeof connector child class18:52
hemna_but that's not even there18:52
openstackgerritWalter A. Boring IV (hemna) proposed openstack/os-brick master: Remove unnecessary NVMEOF connector protocol  https://review.openstack.org/64286018:54
hemna_ok18:54
hemna_I think that gets us back to sanity, unless I'm missing something else18:55
eharneyi didn't follow the "before the release went out" bit, wasn't this already all in os-brick 2.8.0?18:57
hemna_grr18:58
whoami-rajathemna_: why did they even have a protocol mapping when the scaleio protocol was being used. :/18:58
whoami-rajatI think it's better to ask the driver maintainers the reason for this unnecessary(or not) code added/changed18:58
hemna_so we did a release of os-brick with this stuff in it ?18:58
hemna_eharney:I was hoping that mess hadn't made it into a drop to pypi yet18:58
hemna_but nobody is using it either way at this point, so it's probably ok.18:59
* hemna_ ducks18:59
hemna_too many fires....too little time18:59
whoami-rajathemna_: yes that's why we will be doing FFE19:00
eharneyhemna_: whoami-rajat: so the spdk driver uses storage_protocol "NVME-oF"19:05
eharneywas that supposed to be the user of nvmeof?19:06
jungleboyjhemna_:  If the connectors have been released but we don't have any drivers merged using them, then yes it should be safe to revert things.  Right?19:18
hemna_jungleboyj:yah I believe so19:22
jungleboyjCool.19:22
eharneyso, i asked a question there about that ^19:22
jungleboyjeharney:  About the spdk driver?19:23
hemna_oh the spdk.py uses a target driver19:23
hemna_looking19:23
hemna_hrmm19:24
hemna_so, who uses 'nvme' ?19:24
whoami-rajathemna_: the initiator name used in spdk is still nvme19:25
hemna_https://github.com/openstack/cinder/blob/master/cinder/volume/targets/nvmeof.py#L6319:25
jungleboyjwhoami-rajat:  That is what I am seeing in spdk19:25
hemna_whoami-rajat:url ?19:26
eharney???19:26
hemna_I don't see 'nvme' for the connector protocol19:27
hemna_afaik spdk.py uses a target driver19:27
hemna_which results in 'nvmeof'19:27
hemna_quit confusing19:28
hemna_that target driver was added 7 months ago to use 'nvmeof'19:29
hemna_and yet 'nvmeof' was only added 4 months ago to os-brick19:29
* hemna_ scratches head.....19:29
* jungleboyj is so confused at the moment.19:30
hemna_sorry19:30
hemna_the nvvmeof.py target driver was added 2 years ago19:30
hemna_https://github.com/openstack/cinder/blame/master/cinder/volume/targets/nvmeof.py#L2819:31
whoami-rajathemna_: oh its only for copy_image_to_vol and copy_vol_to_image https://github.com/openstack/cinder/blob/master/cinder/volume/drivers/spdk.py#L33719:31
hemna_isn't sure that there was an nvmeof supported connector in os-brick 2 years ago19:31
hemna_https://github.com/openstack/os-brick/blame/master/os_brick/initiator/__init__.py#L6619:32
hemna_4 months ago19:32
hemna_!19:32
hemna_https://github.com/openstack/os-brick/commit/905e73129bea8a141485ef00259c5e40d609bafc19:33
hemna_that was added 2 years ago19:33
hemna_but the protocol was 'nvme'19:33
hemna_not 'nvmeof'19:33
jungleboyjOh, so you are saying that the spdk driver is using NVMEOF because both NVME and NVMEOF map to the same NVMeConnector ?19:33
hemna_so cinder was returning 'nvmeof'19:33
hemna_so this never workd19:33
hemna_worked19:33
hemna_jungleboyj: yup19:33
hemna_cinder has been returning 'nvmeof' (initialize_connection calls -> nova -> os-brick)19:34
hemna_but there hasn't been an 'nvmeof' supported protocol until 4 months ago19:34
hemna_so it was broken for a long time19:34
* jungleboyj sighs19:34
hemna_https://review.openstack.org/#/c/642860/19:35
hemna_so that revert should not merge IMHNO19:35
eharneyhemna_: https://review.openstack.org/#/c/482640/ seems to use nvmeof via brick19:35
eharneymaybe e0ne can fill in some info there19:36
hemna_hrmm yah it hard coded the use of NVME19:37
eharneysorta19:37
eharneyi think nvmeof= in libvirt_volume_drivers means it could consume it as nvmeof...19:37
hemna_yah19:38
hemna_that's how nova finds the nova libvirt volume driver, which in turn uses os-brick19:38
hemna_and internally hard code maps to nvme protocol19:38
hemna_so it worked whenever that patch landed19:38
hemna_10 months ago I suppose19:38
hemna_so I abandoned https://review.openstack.org/#/c/642860/19:39
hemna_but we have 2 protocols pointing to the same connector19:39
hemna_smh19:39
hemna_nova should change it's libvirt volume driver to point to nvmeof19:40
hemna_very confusing19:40
hemna_once we do that, I don't see a need to have the nvme protocol support in os-brick19:40
hemna_unless someone can find an instance where 'nvme' is used19:40
whoami-rajathemna_ https://review.openstack.org/#/c/482640/31/nova/virt/libvirt/volume/nvme.py@35 this is creating a connector with NVME, why we are still keeping NVMEOF in connectors list, could you please explain?19:44
*** thgcorrea has quit IRC20:00
*** e0ne has joined #openstack-cinder20:20
*** dave-mccowan has quit IRC20:24
*** senrique_ has joined #openstack-cinder20:32
*** enriquetaso has quit IRC20:34
*** luizbag has quit IRC20:46
*** gmann_afk is now known as gmann20:54
*** zhubx has quit IRC20:59
*** lseki has quit IRC20:59
*** zhubx has joined #openstack-cinder20:59
*** lseki has joined #openstack-cinder21:00
*** abishop has quit IRC21:01
*** eharney has quit IRC21:03
*** mriedem has quit IRC21:05
*** mriedem has joined #openstack-cinder21:08
*** e0ne has quit IRC21:26
*** senrique__ has joined #openstack-cinder21:32
*** senrique_ has quit IRC21:34
*** senrique__ has quit IRC21:47
*** rcernin has joined #openstack-cinder22:01
*** bal has quit IRC22:05
*** carlos_silva has quit IRC22:09
*** hoonetorg has quit IRC22:09
*** hoonetorg has joined #openstack-cinder22:10
*** whoami-rajat has quit IRC22:12
*** gnufied has quit IRC22:15
openstackgerritSam Morrison proposed openstack/cinder master: Include availability-zone of a pool in get-pools API request  https://review.openstack.org/59986622:23
openstackgerritSam Morrison proposed openstack/cinder master: Include availability-zone of a pool in get-pools API request  https://review.openstack.org/59986622:26
*** enriquetaso has joined #openstack-cinder22:38
*** angela-s has quit IRC22:39
*** ianychoi_ is now known as ianychoi22:47
*** mriedem has quit IRC22:51
*** tkajinam has joined #openstack-cinder22:53
*** alkhodos has joined #openstack-cinder23:02
*** enriquetaso has quit IRC23:07
*** itlinux has joined #openstack-cinder23:37
*** lseki has quit IRC23:58
« Monday, 2019-03-11 Index Wednesday, 2019-03-13 »

Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!