Wednesday, 2020-07-15

*** rmcall has quit IRC00:06
*** knomura has joined #openstack-cinder00:54
*** akrog has joined #openstack-cinder00:58
*** geguileo has quit IRC00:58
*** knomura has quit IRC00:58
openstackgerritwanwencong proposed openstack/cinder master: Avoid rbd import operation to reduce encryption volume creation time
*** xinliang has joined #openstack-cinder01:32
*** spatel has joined #openstack-cinder02:10
*** sapd__x has joined #openstack-cinder02:20
*** rmcall has joined #openstack-cinder02:31
*** spatel has quit IRC02:49
*** rcernin has quit IRC02:56
*** rcernin has joined #openstack-cinder02:57
*** ociuhandu has joined #openstack-cinder03:01
*** ociuhandu has quit IRC03:06
*** brinzhang0 has joined #openstack-cinder03:15
openstackgerritMerged openstack/cinder stable/stein: PowerMax Driver - Snapshot restore payload fix
*** brinzhang_ has quit IRC03:18
*** brinzhang has joined #openstack-cinder03:19
*** spatel has joined #openstack-cinder03:22
*** spatel has quit IRC03:22
*** brinzhang0 has quit IRC03:22
*** spatel has joined #openstack-cinder03:23
*** spatel has quit IRC03:23
*** spatel has joined #openstack-cinder03:30
*** spatel has quit IRC03:30
*** redrobot has quit IRC03:36
*** psachin has joined #openstack-cinder03:38
*** brinzhang_ has joined #openstack-cinder03:41
*** brinzhang has quit IRC03:44
*** xinliang has quit IRC04:01
*** brinzhang0 has joined #openstack-cinder04:03
*** brinzhang_ has quit IRC04:05
*** yaawang has quit IRC04:10
*** markvoelker has joined #openstack-cinder04:11
*** yaawang has joined #openstack-cinder04:11
*** sapd__x has quit IRC04:12
*** markvoelker has quit IRC04:15
openstackgerritKazufumi Noto proposed openstack/cinder master: Support modern compression algorithms in cinder backup
*** sonivi has joined #openstack-cinder04:35
*** vishalmanchanda has joined #openstack-cinder04:50
openstackgerritwanwencong proposed openstack/cinder master: Avoid rbd import operation to reduce encryption volume creation time
*** sonivi has quit IRC04:58
*** sapd__x has joined #openstack-cinder04:58
*** markvoelker has joined #openstack-cinder05:01
*** markvoelker has quit IRC05:05
*** sonivi has joined #openstack-cinder05:15
*** raghavendrat has joined #openstack-cinder05:15
*** brinzhang_ has joined #openstack-cinder05:16
*** yaawang has quit IRC05:16
*** abdysn has joined #openstack-cinder05:16
soniviHi Team, I have python code which consumes third party API05:16
*** yaawang has joined #openstack-cinder05:16
soniviwanted to write unit test cases for python code which consumes API05:16
soniviPlease suggest lib and links to follow05:17
*** brinzhang0 has quit IRC05:18
*** brinzhang0 has joined #openstack-cinder05:21
*** udesale has joined #openstack-cinder05:22
*** brinzhang_ has quit IRC05:24
*** whoami-rajat has joined #openstack-cinder05:34
*** brinzhang_ has joined #openstack-cinder05:40
*** brinzhang0 has quit IRC05:43
*** udesale has quit IRC06:07
raghavendrathi whoami-rajat: are you online ?06:11
whoami-rajatraghavendrat, hey06:11
raghavendratregarding yesterday you provided +206:12
raghavendratlater e0ne also provided +2. i am waiting for +W06:13
whoami-rajatnot sure why he didn't approve06:13
whoami-rajatraghavendrat, done06:13
whoami-rajatraghavendrat, i hope this parameter is tested out in the HPE CI06:14
raghavendratthanks whomai-rajat: for quick approval06:14
*** udesale has joined #openstack-cinder06:20
raghavendrata UT has been added in code review. Let me double check from HPE CI06:20
*** yaawang has quit IRC06:26
*** yaawang has joined #openstack-cinder06:27
*** takamatsu has quit IRC06:59
*** rcernin has quit IRC07:07
*** ociuhandu has joined #openstack-cinder07:17
*** rcernin has joined #openstack-cinder07:17
*** ociuhandu_ has joined #openstack-cinder07:28
*** ociuhandu has quit IRC07:31
*** tosky has joined #openstack-cinder07:46
*** udesale has quit IRC07:47
*** raghavendrat has quit IRC07:47
*** udesale has joined #openstack-cinder07:47
*** elod_off is now known as elod07:51
*** e0ne has joined #openstack-cinder08:06
*** markvoelker has joined #openstack-cinder08:12
*** markvoelker has quit IRC08:17
*** yaawang has quit IRC08:24
*** rcernin has quit IRC08:27
*** yaawang has joined #openstack-cinder08:31
*** sonivi has quit IRC08:32
*** ociuhandu_ has quit IRC08:33
*** ociuhandu has joined #openstack-cinder08:33
*** bhagyashris is now known as bhagyashris|lunc08:43
*** Lucas_Gray has joined #openstack-cinder08:45
*** rcernin has joined #openstack-cinder08:48
*** xinliang has joined #openstack-cinder08:50
openstackgerritIvan Kolodyazhny proposed openstack/cinder master: DNR. Add cinder-plugin-ceph-tempest-mn-aa job
openstackgerritIvan Kolodyazhny proposed openstack/cinder master: DNR. Add cinder-plugin-ceph-tempest-mn-aa job
openstackgerritIvan Kolodyazhny proposed openstack/cinder master: DNR. Add cinder-plugin-ceph-tempest-mn-aa job
*** rcernin has quit IRC09:17
*** martinkennelly has joined #openstack-cinder09:17
openstackgerritLuigi Toscano proposed openstack/cinderlib master: WIP functional tests: directly set PATH
*** xinliang has quit IRC09:27
*** akrog is now known as geguileo09:41
openstackgerritDirk Mueller proposed openstack/cinder master: Switch from unittest2 compat methods to Python 3.x methods
*** yaawang has quit IRC10:24
*** yaawang has joined #openstack-cinder10:25
toskygeguileo: hi! I didn't notice the other nick. We have a problem with cinderlib tests in the lio-lvm-barbican test, have you seen yesterday night discussion?10:32
*** mvkr has quit IRC10:34
*** bhagyashris|lunc is now known as bhagyashris10:37
openstackgerritGorka Eguileor proposed openstack/cinderlib master: RBD: Fix error on disconnect
*** sonivi has joined #openstack-cinder10:48
*** martinkennelly has quit IRC10:51
*** tosky has quit IRC11:05
*** martinkennelly has joined #openstack-cinder11:07
*** rcernin has joined #openstack-cinder11:09
*** tosky has joined #openstack-cinder11:10
geguileotosky: I saw you saying you would ping me today...11:12
geguileotosky: and I saw you proposing a patch to remove the PATH workaround in tox.ini11:12
geguileobut the tests hadn't finished when I checked a bit ago11:13
geguileonot finished yet11:13
*** rcernin has quit IRC11:13
geguileotosky: what I don't get is why LIO barbican is failing...11:13
smcginnisgeguileo: Do you recall why was added, and if there is a tox bug somewhere tracking that?11:14
geguileoI mean, it worked, and now it fails?11:14
toskygeguileo: we were not totally sure yet, the guess was about something related to using the wrong stestr or so11:14
toskysmcginnis: removing the PATH workaround worked, in the way the the failure is exactly the same :)11:15
toskyso it's probably not the same issue11:15
smcginnisAh, ok. At least that eliminates one possibility.11:15
rosmaitatosky: so it still can't find ddt?11:15
toskyrosmaita: it can't11:15
toskygeguileo: the discussion is here:
smcginnisSo just need to figure out why the system stestr is being called and not the tox stestr.11:15
geguileosmcginnis: I remember some commands not being found because only the ones in the venv were found11:16
geguileosmcginnis: I think it was for the Ceph backend11:16
geguileoIn theory I appended the system paths so that venv paths had priority11:16
smcginnisGot it.11:17
toskysmcginnis: it's worth noting that the functional tox environment is an open one (use the system packages), so stestr seems to be correctly used from /usr/local11:17
geguileotosky: thanks11:17
toskygeguileo: the venv path was already there technically (~/.tox/functional)11:17
*** martinkennelly has quit IRC11:18
toskybut again the tox environment has sitepackages=True, so that doesn't matter11:18
smcginnisMaybe because stestr is in whitelist_externals?11:19
smcginnisOr I suppose we could be explicit and change the command to "{envbindir}/stestr"11:21
toskysmcginnis: I tried to remove it from whitelist_externals in the last patch stuck in the queue11:21
toskybut I suspect {envbindir} will still point to the external stestr, because it's already found there thanks to sitepackages=True11:22
*** sonivi has quit IRC11:22
toskyI mean, it should just work™11:22
*** sapd__x has quit IRC11:22
smcginnis{envbindir} shouldn't find the site package first. That would be explictly pointing to the one in the venv.11:23
openstackgerritIvan Pchelintsev proposed openstack/cinder master: Add Cinder driver for Dell EMC PowerStore
rosmaitatox docs say when sitepackages=true, "In cases where a command line tool is also installed globally you have to make sure that you use the tool installed in the virtualenv by using python -m <command line tool> (if supported by the tool) or {envbindir}/<command line tool>."11:24
smcginnisAh, well there you go.11:24
toskybut there is no stestr in the venv with sitepackages=True!11:24
smcginnisDon't we have stestr in our test-requirements?11:25
smcginnisThen it should be there.11:25
smcginnisOtherwise we've got bigger issues.11:26
smcginnisJust ran "tox -e functional --notest" and confirmed stestr is installed in the venv.11:29
*** rcernin has joined #openstack-cinder11:31
geguileosmcginnis: but you don't have stestr out of the venv, right?11:31
openstackgerritWalter A. Boring IV (hemna) proposed openstack/cinder master: Add ceph iscsi volume driver
*** udesale_ has joined #openstack-cinder11:31
*** Lucas_Gray has quit IRC11:32
smcginnisgeguileo: I do not have it globally installed, no. But apparently the test node does. The failure we saw indicated it installed our packages in the venv, but then failed to find ddt because it was using the system installed stestr instead of the one in our venv.11:32
geguileosmcginnis: that's because using sitepackages it won't install stestr in the venv since it's already present on the host11:33
smcginnisSo based on what rosmaita found in the docs, it does appear the issue is we need to use {envbindir}/stestr since we are using sitepackages=True11:33
geguileosmcginnis: but I believe it won't work even with that11:33
geguileobecause it won't be installed in the venv (like tosky was saying)11:33
smcginnisI thought it would still install it. Just that setting would allow the use of packages outside of the venv.11:34
*** udesale has quit IRC11:34
geguileoafaik it won't install packages that are already in the host11:34
geguileobut shouldn't stestr be smart enough to know it's running on a venv11:34
smcginnisI saw it getting installed in the venv, but that was the tempest one.11:37
smcginnisUnder the functional test run: Requirement already satisfied: stestr>=1.0.0 in /usr/local/lib/python3.6/dist-packages11:37
smcginnisSo I think if we did "{envbindir}/python -m pip install stestr" we could work around that.11:37
geguileoI think this is an stestr bug11:37
smcginnisHow so?11:37
geguileobecause it's not using the venv python executable11:38
smcginnisNot sure it can since we are calling an executable outside of our venv.11:38
jkulikif it's installed via debian package, then it probably got patched to use /usr/bin/python3 instead of /usr/bin/env python311:39
geguileosmcginnis: it can check VIRTUAL_ENV11:39
smcginnisThanks jkulik, that's probably it.11:39
smcginnisgeguileo: I guess it could. But I don't think that's normal for every python command to do its own check for running in a virtual env.11:40
geguileosmcginnis: you are probably right on that one11:41
openstackgerritWalter A. Boring IV (hemna) proposed openstack/cinder master: [vmware] ensure datastores exist while fetching stats
openstackgerritWalter A. Boring IV (hemna) proposed openstack/cinder master: [vmware] validate storage profile at startup
geguileooooooooh, I think I found the problem!!  going to check now11:43
smcginnisI wonder if install_command should actually be added to the functional environment.11:43
smcginnisBasically the same as the default install command defined in [testenv], but instead of "pip install" do "{envbindir}/pip install"11:44
geguileosmcginnis: tosky found the problem11:44
* smcginnis waits to see the patch11:45
geguileoor not...11:45
*** Lucas_Gray has joined #openstack-cinder11:46
*** rcernin has quit IRC11:47
hemnageguileo ping11:47
geguileoyup, that's the problem11:47
geguileosmcginnis: tosky the problem is that the stestr file in the host has hardcoded the location of the python binary11:48
geguileosmcginnis: tosky instead of using /usr/bin/env python it uses #!/usr/bin/python3.711:48
geguileosmcginnis: tosky so it doesn't use the virtual environment  :-(11:48
smcginnisAs jkulik pointed out. :)11:48
geguileosmcginnis: oh, I missed that  XDXD11:49
hemnawould it be ok to instead add the az to the request_spec['availability_zones'] list as a single entry if the az isn't in the type?11:49
hemnaI presume of the volume type has a list of az's in it, then don't add the volume's az to the list, since the volume type's list of az's should be used instead.11:50
openstackgerritSean McGinnis proposed openstack/cinderlib master: WIP: Testing installing site packages in vevn
rosmaitasmcginnis: i think i must have misread that warning:
smcginnisrosmaita: I think it still applies. Since we have stestr in whitelist_externals, we don't get that same warning message. I think.11:53
geguileohemna: That would add unnecessary code in my opinion11:54
toskyjkulik, smcginnis, geguileo : but if it's installed in /usr/local, it's not the Debian package11:54
geguileohemna: if you think that the request_spec should be fully formed at the API level you can set 'resource_properties': volume11:54
hemnaand copy the entire volume properties in there?11:55
hemnathat seemed to be way overkill just to get the az11:55
hemnathat's worse than what I'm doing already I think.11:55
geguileohemna: you don't need to copy anything11:55
geguileohemna: you just set the reference in the API11:55
geguileothough when serializing it will get copied11:56
geguileobut the concept is that those 2 keys should contain the same values11:56
hemnayup, since it's sent over rpc, it will get copied and serialized11:56
geguileothat's why I was saying to do it on the other end11:56
geguileoafter all it's not new data, it's just that we need it in both places for backward compatibility11:57
hemnaso what I'm doing already is more efficient, since that's all it needs11:57
hemnathen later we can migrate to the ovo and be done with it11:57
geguileohemna: but if there's another driver that uses resource_properties we'll have another bug11:57
hemnabut this will be backwards compatible/backportable for now11:57
geguileohemna: and if there isn't another driver we can just fix that driver11:57
geguileothat wasn't updated when we imported it from oslo.incubator11:57
geguileotosky: 'i have checked my devstack and my stestr is in /usr/local/bin/stestr11:58
geguileotosky: and it has #!/usr/bin/python3.711:58
*** Lucas_Gray has quit IRC11:58
hemnaI was a bit afraid to touch the azfilter since it's been around forever11:58
*** Lucas_Gray has joined #openstack-cinder11:59
geguileohemna: that's reasonable, then we can just add 'resource_properties': request_spec['volume_properties']11:59
geguileoon the scheduler reception of the request11:59
hemnathat seems hackish to me11:59
geguileorequest_spec['resource_properties'] = request_spec['volume_properties']11:59
hemnainstead of passing the az in the request spec12:00
hemnaaltering the request in the scheduler :(12:00
geguileowe are already doing it12:00
hemnayah I saw that, not that it seems right to do12:00
geguileoand you only alter it to add backward compatibility12:00
geguileoI'm more inclined to just fix the filter that we didn't fix in 2015 when we imported the filter12:01
hemnamakes me think we have a mess on our hands if the scheduler is altering requests coming from the api12:01
geguileoyeah, because we didn't import the filter as we should12:01
toskygeguileo: but shouldn't we see the same issue all around? tempest runs from a venv12:01
rosmaitageguileo: tosky: why do we need sitepackages=true for the functional testenv?12:01
geguileohemna: and then we hacked our way around it12:01
geguileohemna: so yes, it's a hack, but yours is one as well12:01
geguileohemna: In my mind the right approach would be to fix the filter, second best to set resource to volume12:02
hemnanot sure I agree that this is a hack, as it adds a missing property to the request spec12:02
geguileotosky: Cinder set sitepackages to false12:02
geguileohemna: the missing property is the whole resource_properties12:03
geguileohemna: not a single field in that key12:03
hemnathe azfilter is looking for 2 places currently for the az, spec.get('availability_zones') and resource_properties['availability_zone']12:03
geguileorosmaita: iirc it's for ceph12:03
geguileorosmaita: because we install ceph-common on the host which provides ceph libraries12:03
rosmaitathat's what i was thinking12:03
geguileorosmaita: and the only existing ceph libraries in PyPi are outdated and fail to load12:04
toskyin a venv I installed tempest and the stestr's shebang points to the python inside the venv12:04
*** Lucas_Gray has quit IRC12:04
rosmaitatoo bad sitepackages isn't (True|False|Maybe)12:04
geguileohemna: it's loooking at resource_properties because we messed the import of that filter12:04
geguileotosky: but that's because you installed it in the venv12:05
geguileotosky: but if your system had stestr in the host12:05
hemnarosmaita, I'm not sure why but, it seems that ceph doesn't put some of their libs even in pypi.  they have them in github and leave them there for distros to create packages for them12:05
geguileotosky: then the venv wouldn't install it12:05
hemnaceph-iscsi is one, tcmu-runner12:05
geguileohemna: so true, they don't push to PyPi  :-(12:05
hemnaceph-iscsi is pure python, and could go in pypi12:05
hemnageguileo, so the more long term fix for this is what?   we look into the request_spec['volume'] ovo for the fields ?12:06
hemnayou would think that az would be in the same spot accross all API request_specs12:07
rosmaitatosky: so what do we do? leave sitepackages=true but pip-install the stestr we want into the virtualenv?12:07
geguileohemna: the long term fix is using OVOs, and for this particular bug we can either make sure "resource_properties" is set, or fix the filter itself12:08
geguileorosmaita: I think that's probably the easiest solution12:08
geguileorosmaita: though I think we'll have to force install it12:08
hemnaok that's what my patch does...adds the az to the resource_properties.  :)12:08
geguileohemna: but we must add the whole resource_properties, not just a single field12:08
hemnaok so add the entire volume entry then12:09
geguileohemna: because resource_properties and volume_properties are the same thing12:09
geguileohemna: but maybe I'm the only one thinking this way12:09
toskyrosmaita, geguileo: or try to use `python3 stestr...` explicitly?12:09
toskyrosmaita, geguileo: or python -m ... as suggested by smcginnis ?12:10
geguileohemna: I'm not going to veto the patch, it's just my opinion12:10
toskyso that the right version of python would be used12:10
rosmaitatosky: i think you are right that if sitepackages=True, stestr is not installed in the virtualenv12:10
geguileotosky: smcginnis's python -m could work and it's a lot cleaner12:10
toskythe forced interpreter should override the default from shebang12:10
geguileotosky: give me a sec to confirm it works12:10
geguileotosky: the -m works12:13
geguileotosky: it directly bypassed the whole /usr/local/bin/stestr file and goes directly to stestr.cli.main12:14
toskyas expected, at least this time12:14
geguileoso sys.executable points to the right python interpreter12:14
geguileoI think that's the nicest solution12:14
toskydo we need a separate patch, or can we fold it inside ?12:15
geguileotosky: I'm ok to include it in that one12:15
rosmaitatosky: make sure you have a comment about doing this because of the debian stestr distro12:16
*** baojg has quit IRC12:17
toskyrosmaita: I'm still not sure why the executable under /usr/local is related to debian packages12:17
*** baojg has joined #openstack-cinder12:17
toskyit should be under /usr/bin; I think that's the stestr installed by devstack12:19
tosky2020-07-14 22:15:00.850603 | controller |   Downloading (117 kB)12:19
rosmaitatosky: ok, well be cagy about it ... say something about we want to call the module directly because some scripts in some distros may not use the env python and we are using sitepackages12:19
tosky"installed outside the venv"12:21
toskygeguileo: should I also keep the removal of bash -i -c ? I think it makes sense to remove it, see
*** enriquetaso has joined #openstack-cinder12:23
*** tkajinam has quit IRC12:24
*** martinkennelly has joined #openstack-cinder12:26
*** Lucas_Gray has joined #openstack-cinder12:30
*** vapjes has joined #openstack-cinder12:35
*** sonivi has joined #openstack-cinder12:49
*** martinkennelly has quit IRC12:53
soniviHi, I want to write a unit test cases for python code which consumes third party API12:53
soniviAny suggestion are welcome on which lib/link to use12:54
rosmaitasonivi: how about some context? is this for a new driver or something?12:55
soniviyes its new driver but not for OpenStack.12:56
*** whoami-rajat has quit IRC12:56
rosmaitasonivi: some of our drivers use third party APIs, you could see what they do12:56
soniviits similar to what we have in cinder12:56
sonivirosmaita: yes.. thats good idea12:57
rosmaitalook for drivers that have a rest client12:57
soniviI have seen one client code HPE3parclient code, in which flask is used to host the APIs which returns the expected response12:58
*** whoami-rajat has joined #openstack-cinder12:58
hemnasonivi I wrote that one12:58
soniviyes walter, even I was a part of it later to you12:59
hemnathe flask server is just help test the client calls12:59
openstackgerritIvan Kolodyazhny proposed openstack/cinder master: DNR. Add cinder-plugin-ceph-tempest-mn-aa job
hemnaso it didn't have to talk to an actual 3par on the backend for CI.13:00
hemnaflask worked well for that13:00
sonivihemna: yes flask worked well for it13:00
rosmaitae0ne: "Do Not Resuscitate" ?13:01
e0nerosmaita: it's a good option :13:01
rosmaitathat's my plan for stable/pike!13:01
*** whoami-rajat has quit IRC13:03
*** whoami-rajat has joined #openstack-cinder13:04
hemnageguileo so, looking at the request_spec for volume create.   it looks like the api flow creates a request spec with 'availability_zones' which is a list of 1 entry in this case for create13:07
hemnathe azfilter looks for availability_zones13:07
sonivihemna: from where these variable is coming from13:08
hemnaI kinda think that retype should use that13:08
hemnasonivi shouldn't be using nose anymore. I created a PR for migrating away from that last year.
hemnathere are other PR's up there that should be merged as well, from geguileo13:11
openstackgerritSean McGinnis proposed openstack/cinderlib master: WIP: Testing running stestr through venv python
sonivi<hemna>:ok..I am not aware much on tox. I will go through it13:16
toskysmcginnis: did you remove the  stestr run --serial {posargs}" on purpose?13:16
smcginnisHaha, oops. Nope.13:16
openstackgerritSean McGinnis proposed openstack/cinderlib master: WIP: Testing running stestr through venv python
*** brinzhang_ has quit IRC13:17
*** whoami-rajat has quit IRC13:25
*** whoami-rajat has joined #openstack-cinder13:25
*** baojg has quit IRC13:33
*** sonivi has quit IRC13:42
*** Guest14648 has joined #openstack-cinder13:43
*** rcernin has joined #openstack-cinder13:44
*** Guest14648 is now known as redrobot13:45
*** rcernin has quit IRC13:48
openstackgerritLuigi Toscano proposed openstack/cinderlib master: Fix and improve cross-project functional testing
rosmaitaCourtesy reminder: Cinder meeting in #openstack-meeting-alt at 1400 UTC13:58
rosmaitajungleboyj rosmaita smcginnis tosky whoami-rajat m5z e0ne geguileo eharney walshh_ jbernard lseki sfernand rajinir enriquetaso ^^13:58
hemnaman the more I look into the filtering and request specs, it's really a mess13:58
whoami-rajatrosmaita, thanks13:59
*** Liang__ has joined #openstack-cinder13:59
*** Liang__ is now known as LiangFang13:59
hemnalots of duplicated data, some data in one place, other data in other places, some filters use 1 location, another filter looks in a separate location...14:00
*** sfernand has joined #openstack-cinder14:04
*** whoami-rajat has quit IRC14:04
geguileohemna: sooooooo true14:05
geguileoit's a bit of a mess14:05
*** whoami-rajat__ has joined #openstack-cinder14:06
*** whoami-rajat___ has joined #openstack-cinder14:07
*** abdysn has quit IRC14:26
hemnathe base of the filter_properties is automatically added to every request to the scheduler basically here:14:41
hemnawhich contains duplicated data from the request_spec which is also added to the filter_properties at the API level14:43
hemnajust smells like the request_spec should be a separate dict from the filter_properties14:43
*** dave-mccowan has quit IRC14:43
hemnaso for retype, I think the API should just populate the request_spec['availability_zones'] if they don't exist in the volume type14:47
*** dave-mccowan has joined #openstack-cinder14:47
hemnageguileo like this:
hemnaso I think that'd work:   if the volume type you are changing to has AZ's set, it will use those to filter against the backends, else it will use the existing volume's az14:50
hemnaso if someone wants to force an AZ(s) on a volume type, they still can by setting them in the type extra specs.14:51
geguileohemna: I think that would replace the one from the volume_type14:51
hemnaif the volume type has the az set, it will use those14:52
hemnahence the else14:52
geguileooh, sorry14:52
geguileoI'm ok with fixing this bug like this14:53
geguileoand I'll propose a patch to fix the AZ filter14:53
hemnaok I'll update my patch with that14:53
hemnaat some time we should revisit how the request_spec and filter_properties are used and populated14:54
hemnait's a mess14:54
geguileohemna: because AZ cannot be None, right?14:54
hemnacorrect.   The AZ should be part of the request_spec for retype14:54
hemnaotherwise you can get a volume being retyped to an AZ you don't want.14:54
hemnaand won't find a backend to take the request, which is the original bug I found that this patch is trying to fix.14:55
hemnait basically picks a backend in another az that can't accept the request, and then retype pukes.14:55
hemnaa retype should stay in the same AZ unless the type specifies an AZ14:56
geguileowhat I was asking is if we could have a volume without an AZ, and looking at the code we won't14:57
hemnaI think there is always an AZ set14:59
smcginnisLooks like actual bandit run time is about 3 minutes. The whole job is just over 7 usually. So more than 50% overhead running it as a separate job.15:00
smcginnisBut then, I doubt it is useful at all on stable branches.15:01
openstackgerritWalter A. Boring IV (hemna) proposed openstack/cinder master: Fix volume retype with AZ
*** knomura has joined #openstack-cinder15:02
eharneyfyi, i was chasing around lio-barbican failures and saw some weird things where it looks like nova-compute gets stuck for a bit.  experimenting with now but it's kind of a shot in the dark15:02
rosmaitaeharney: thanks for looking into that15:03
eharneyi think the service gets blocked up for 90-100+ seconds and then starts timing out amqp requests, but i'm not really sure why15:04
LiangFangI encountered failure of this job, but recheck passed15:04
LiangFangrosmaita: Thanks rosmaita and Jay reviewed volume local cache patches, I have updated the patches this week. os-brick cinder15:05
eharneytosky: i didn't find anything in the logs that pointed to any obvious resource shortage, maybe will look for that specifically again15:05
eharneytosky: ooh....   in the mysql log:  2020-07-14T17:19:02.735148Z 0 [Note] InnoDB: page_cleaner: 1000ms intended loop took 125377ms. The settings might not be optimal. (flushed=201 and evicted=0, during the time.)15:07
eharney(at the time of failure)15:07
rosmaitaLiangFang: will take a look at your revisions today15:10
*** LiangFang has quit IRC15:10
hemnageguileo thanks for the help man15:10
rosmaitaeharney: i like that error message -- taking 125x longer than it should does indeed indicate that something sub-optimal is going on15:12
eharneyrosmaita: it looks like there are a handful of knobs we can turn in mysql to remedy this kind of problem, may take a swing at a devstack patch and see what happens15:13
eharneymysql also reported an error "writing communication packets" back to nova_cell1 with that error, so it fits15:14
rosmaitathat is really promising15:14
geguileohemna: thank you for your patience and for fixing the issue :-)15:14
*** whoami-rajat__ has quit IRC15:15
*** whoami-rajat___ is now known as whoami-rajat__15:15
openstackgerritEric Harney proposed openstack/cinder-tempest-plugin master: WIP: Test workaround for lio-barbican failures
*** hamalq has joined #openstack-cinder15:31
*** hamalq_ has joined #openstack-cinder15:34
*** hamalq has quit IRC15:38
*** martinkennelly has joined #openstack-cinder15:40
openstackgerritBrian Rosmaita proposed openstack/cinder stable/ussuri: Move non-voting jobs to experimental
*** rosmaita has left #openstack-cinder15:54
*** rosmaita has joined #openstack-cinder15:57
*** baojg has joined #openstack-cinder16:28
*** ociuhandu_ has joined #openstack-cinder16:36
*** udesale_ has quit IRC16:38
*** ociuhandu has quit IRC16:39
*** ociuhandu_ has quit IRC16:40
*** e0ne has quit IRC16:46
*** e0ne has joined #openstack-cinder16:47
*** baojg has quit IRC16:49
*** baojg has joined #openstack-cinder16:50
*** psachin has quit IRC16:58
openstackgerritIvan Pchelintsev proposed openstack/cinder master: Add Cinder driver for Dell EMC PowerStore
*** senrique_ has joined #openstack-cinder17:22
*** enriquetaso has quit IRC17:24
*** Lucas_Gray has quit IRC17:41
*** geguileo has quit IRC17:41
openstackgerritIvan Pchelintsev proposed openstack/cinder master: Add Cinder driver for Dell EMC PowerStore
*** geguileo has joined #openstack-cinder17:42
*** e0ne has quit IRC18:00
*** senrique__ has joined #openstack-cinder18:05
*** senrique_ has quit IRC18:07
*** martinkennelly has quit IRC18:08
*** larsks has joined #openstack-cinder18:14
*** vishalmanchanda has quit IRC18:20
*** tosky has quit IRC18:25
*** tosky has joined #openstack-cinder18:25
*** knomura has quit IRC18:25
*** knomura has joined #openstack-cinder18:27
*** knomura has quit IRC18:32
*** baojg has quit IRC18:38
*** baojg has joined #openstack-cinder18:39
*** ociuhandu has joined #openstack-cinder18:46
*** ociuhandu has quit IRC18:50
*** jdillaman has joined #openstack-cinder18:53
*** KeithMnemonic has joined #openstack-cinder19:12
*** rmcall has quit IRC19:22
*** rmcall has joined #openstack-cinder19:23
*** rmcallis has joined #openstack-cinder19:25
*** rmcall has quit IRC19:28
*** rmcallis__ has joined #openstack-cinder19:28
*** rmcallis has quit IRC19:31
toskyit passed!19:48
rosmaitajust saw taht19:53
rosmaitaok, soon as that merges we can have a recheck party19:54
toskythanks all for the suggestions!19:56
toskynow we need to work on devstack-plugin-nfs :)19:57
*** sfernand has quit IRC20:13
*** senrique__ has quit IRC20:35
*** hamalq_ has quit IRC20:46
*** hamalq has joined #openstack-cinder20:48
*** hamalq has quit IRC20:52
*** hamalq has joined #openstack-cinder20:52
*** rmcall has joined #openstack-cinder21:27
*** rmcallis__ has quit IRC21:28
*** rmcall has quit IRC21:30
*** rmcall has joined #openstack-cinder21:31
*** rmcall has quit IRC21:35
*** rmcallis has joined #openstack-cinder21:36
*** rmcallis has quit IRC21:42
openstackgerritLucio Seki proposed openstack/cinder master: [NetApp] Adding support for Adaptive QoS in NetApp driver.
*** logan- has quit IRC22:17
*** rcernin has joined #openstack-cinder22:18
*** logan- has joined #openstack-cinder22:19
*** rcernin has quit IRC22:20
*** rcernin has joined #openstack-cinder22:21
*** spatel has joined #openstack-cinder22:29
*** vapjes has quit IRC22:33
*** spatel has quit IRC22:34
*** hamalq has quit IRC22:35
openstackgerritMerged openstack/cinderlib master: Fix and improve cross-project functional testing
*** baojg has quit IRC22:43
*** baojg has joined #openstack-cinder22:44
*** tosky has quit IRC22:48
openstackgerritMerged openstack/cinder-specs master: Support modern compression algorithms in cinder backup
*** tkajinam has joined #openstack-cinder22:55
*** tkajinam has quit IRC22:55
*** tkajinam has joined #openstack-cinder22:58
*** markvoelker has joined #openstack-cinder23:11
*** markvoelker has quit IRC23:15
*** knomura has joined #openstack-cinder23:38
*** knomura has quit IRC23:42
*** baojg has quit IRC23:58
*** baojg has joined #openstack-cinder23:59

Generated by 2.17.2 by Marius Gedminas - find it at!