Tuesday, 2017-07-18

« Monday, 2017-07-17 Index Wednesday, 2017-07-19 »
*** gcb has quit IRC00:53
*** harlowja has quit IRC01:34
openstackgerritzhongjun proposed openstack/manila master: Add share usage size tracking  https://review.openstack.org/46505501:45
openstackgerritOpenStack Proposal Bot proposed openstack/manila master: Updated from global requirements  https://review.openstack.org/48455601:48
openstackgerritOpenStack Proposal Bot proposed openstack/python-manilaclient master: Updated from global requirements  https://review.openstack.org/48341901:56
*** ameade_ has joined #openstack-manila02:06
*** scottda_ has joined #openstack-manila02:07
*** toabctl_ has joined #openstack-manila02:13
*** scottda has quit IRC02:13
*** toabctl has quit IRC02:13
*** assassin has quit IRC02:13
*** ameade has quit IRC02:13
*** scottda_ is now known as scottda02:13
*** toabctl_ is now known as toabctl02:14
*** ameade_ is now known as ameade02:14
*** rhagarty has quit IRC02:16
*** assassin has joined #openstack-manila02:19
*** gcb has joined #openstack-manila02:19
*** rhagarty has joined #openstack-manila02:19
*** assassin is now known as Guest8387402:20
*** chlong has quit IRC03:11
*** harlowja has joined #openstack-manila03:40
*** tuanluong has joined #openstack-manila03:46
openstackgerritTuan Luong-Anh proposed openstack/manila master: Replace test.attr with decorators.attr  https://review.openstack.org/48461603:56
*** markstur has quit IRC04:10
*** rejy has joined #openstack-manila04:13
*** markstur has joined #openstack-manila04:20
*** markstur has quit IRC04:25
*** markstur has joined #openstack-manila04:41
*** markstur has quit IRC04:45
*** harlowja has quit IRC04:54
*** dsariel_ has joined #openstack-manila05:11
*** dsariel has joined #openstack-manila05:11
*** arnewiebalck_ has joined #openstack-manila05:13
*** arnewiebalck has quit IRC05:15
*** arnewiebalck_ is now known as arnewiebalck05:15
*** markstur has joined #openstack-manila05:22
*** gouthamr has quit IRC05:25
*** markstur has quit IRC05:27
*** jprovazn has joined #openstack-manila05:27
*** markstur has joined #openstack-manila06:04
*** markstur has quit IRC06:08
*** lpetrut_ has joined #openstack-manila06:11
*** markstur has joined #openstack-manila06:45
*** markstur has quit IRC06:51
*** lpetrut_ has quit IRC07:02
*** Guest83874 is now known as assassin07:08
*** assassin is now known as Guest109607:08
*** lpetrut_ has joined #openstack-manila07:17
*** lpetrut_ has quit IRC07:29
*** markstur has joined #openstack-manila07:33
*** markstur has quit IRC07:37
*** gcb has quit IRC08:09
*** markstur has joined #openstack-manila08:10
*** Guest1096 has quit IRC08:11
*** assassin has joined #openstack-manila08:11
*** assassin has quit IRC08:11
*** assassin has joined #openstack-manila08:11
*** markstur has quit IRC08:14
*** markstur has joined #openstack-manila08:46
*** markstur has quit IRC08:51
*** ociuhandu has quit IRC09:08
*** markstur has joined #openstack-manila09:23
openstackgerritMayur Indalkar proposed openstack/manila master: Implementation of Manila driver for Veritas Access  https://review.openstack.org/47219009:27
*** markstur has quit IRC09:28
openstackgerritzhongjun proposed openstack/manila master: Fix the exact filter can be filter by inexact value  https://review.openstack.org/48469609:39
*** markstur has joined #openstack-manila10:00
*** markstur has quit IRC10:05
*** gcb has joined #openstack-manila10:11
openstackgerritjunbo.li proposed openstack/manila master: Enable some off-by-default checks  https://review.openstack.org/48470710:13
*** lpetrut_ has joined #openstack-manila10:23
*** markstur has joined #openstack-manila10:37
*** markstur has quit IRC10:42
*** ociuhandu has joined #openstack-manila10:49
*** dsariel_ has quit IRC10:50
*** dsariel has quit IRC10:50
*** rejy has quit IRC10:57
*** markstur has joined #openstack-manila11:14
*** markstur has quit IRC11:18
*** sapcc-bot has joined #openstack-manila11:30
*** sapcc-bot has quit IRC11:30
*** sapcc-bot has joined #openstack-manila11:31
*** dgonzalez_ has joined #openstack-manila11:31
*** carthaca_ has joined #openstack-manila11:31
*** david_1 has joined #openstack-manila11:31
*** dgonzalez_ has quit IRC11:33
*** carthaca_ has quit IRC11:33
*** david_1 has quit IRC11:33
*** lpetrut__ has joined #openstack-manila11:35
*** lpetrut_ has quit IRC11:37
*** markstur has joined #openstack-manila11:51
*** markstur has quit IRC11:55
*** chlong has joined #openstack-manila12:00
*** dustins has joined #openstack-manila12:01
*** jmlowe_ has quit IRC12:04
*** tuanluong has quit IRC12:15
zhongjuntbarron: ping12:25
*** rraja has joined #openstack-manila12:27
*** markstur has joined #openstack-manila12:28
*** ganso has joined #openstack-manila12:28
*** MVenesio has joined #openstack-manila12:29
*** dsariel_ has joined #openstack-manila12:32
*** markstur has quit IRC12:32
*** dsariel has joined #openstack-manila12:33
*** jmlowe has joined #openstack-manila12:38
*** catintheroof has joined #openstack-manila12:51
*** catintheroof has quit IRC12:58
*** markstur has joined #openstack-manila13:04
*** catintheroof has joined #openstack-manila13:06
*** eharney has joined #openstack-manila13:06
*** markstur has quit IRC13:09
*** catintheroof has quit IRC13:11
tbarronzhongjun: pong (just got out of a meeting and saw your ping)13:16
*** dustins has quit IRC13:22
*** dustins has joined #openstack-manila13:25
*** markstur has joined #openstack-manila13:41
*** ociuhandu has quit IRC13:46
*** markstur has quit IRC13:46
*** catintheroof has joined #openstack-manila13:51
*** gouthamr has joined #openstack-manila13:56
*** dustins_ has joined #openstack-manila13:58
*** dustins_ has quit IRC13:58
*** markstur has joined #openstack-manila14:18
*** eharney has quit IRC14:21
*** markstur has quit IRC14:22
vponomaryovbswartz: ping14:25
bswartzpong14:25
vponomaryovbswartz: commit https://review.openstack.org/#/c/312321/ that is being merged has broken tempest tests14:25
bswartzvponomaryov: it passed14:25
bswartzwhat am I missing?14:25
vponomaryovbswartz: under "broken" I mean they do not run14:25
vponomaryovhttps://review.openstack.org/#/c/312321/80/manila_tempest_tests/tests/api/test_rules.py14:26
vponomaryovtest_create_delete_access_rules_with_one_ip and test_create_delete_access_rule_with_cidr didn't run14:26
vponomaryovsee logs: http://logs.openstack.org/21/312321/80/check/gate-manila-tempest-dsvm-mysql-generic-ubuntu-xenial-nv/0cdfa00/console.html14:26
vponomaryovhere is normal logs WITH these tests from another commit: http://logs.openstack.org/56/484556/1/check/gate-manila-tempest-dsvm-mysql-generic-ubuntu-xenial-nv/8225c19/console.html#_2017-07-18_11_34_51_67819314:27
vponomaryovwere 59, became 4914:27
bswartzI see zero failed in both cases14:27
vponomaryovthey do not run14:27
bswartzskipped?14:28
bswartzor disabled?14:28
vponomaryovI guess something wrong data provisioning with "ddt"14:28
bswartzlet's fix the problem as a bugfix14:28
vponomaryovs/data/with data/14:28
vponomaryovbswartz: it means feature is not tested14:29
bswartzidk is zhongjun is still awake, but maybe we can tell her what she needs to fix14:29
bswartzwe've all tested it manually though14:29
vponomaryovbswartz: and we merge it not testing14:29
bswartzwe just need to fix the automation14:29
zhongjunhi, I am awake14:30
bswartzzhongjun: vponomaryov has concerns about skipped tempest tests related to IPv6 access14:31
bswartz#link https://review.openstack.org/#/c/312321/14:31
zhongjunbswartz: I am looking14:31
openstackgerritBen Swartzlander proposed openstack/python-manilaclient master: Enable IPv6 in manila(client)  https://review.openstack.org/32893214:35
*** eharney has joined #openstack-manila14:37
*** rraja has quit IRC14:42
zhongjunvponomaryov: Do you have any idea why it not run the tempest test : test_create_delete_access_rule_with_cidr...14:43
zhongjunvponomaryov:  https://review.openstack.org/#/c/312321/33/manila_tempest_tests/tests/api/test_rules_negative.py@11614:43
vponomaryovzhongjun: you use "chain"14:44
vponomaryovzhongjun: and diswrap it with "*"14:44
vponomaryovzhongjun: so, you depend on its return value14:44
vponomaryovzhongjun: and its possibility to handle "*"14:45
vponomaryovzhongjun: so, I guess, you get empty data there14:45
vponomaryovzhongjun: and tests just do not run14:45
vponomaryovzhongjun: also, you enabled ipv6 for all backends doing so14:46
vponomaryovzhongjun: without possibility to disable ipv6 rules14:46
vponomaryovso, tempest part should be reverted for sure14:47
vponomaryovand reworked14:47
openstackgerritMerged openstack/manila master: Enable IPv6 in manila(allow access)  https://review.openstack.org/31232114:50
*** markstur has joined #openstack-manila14:55
zhongjunvponmaryov: sure,  I will put a new patch to fix it.14:55
zhongjunvponmaryov: since we discussed before, we agree with that we don't need to add ipv6 rules flag  right?  We will support ipv6 access in manila?14:57
zhongjunvponmaryov: I am not sure my memory is right14:58
*** markstur has quit IRC14:59
bswartzzhongjun: what is the question?15:00
bswartzwe should not have a tempest flag for ipv6 tests15:00
bswartzthey should just pass15:00
zhongjunbswartz:Do we need to add flag in tempest to disable ipv6 rules15:01
vponomaryovbswartz: it depends on backend15:01
bswartzI disagree15:01
bswartzthis was an important topic we discussed in barcelona15:01
bswartzipv6 rules are always allowed -- whether the backend supports them or not15:02
bswartzbackends that don't support ipv6 simply ignore the rules15:02
bswartzwhich is no different from a backend that does support ipv6 rules but doesn't have an ipv6 export location configured15:02
bswartzfrom the end user's perspective, the 2 scenarios should look the same15:03
*** markstur has joined #openstack-manila15:03
*** markstur has quit IRC15:03
vponomaryovbswartz: I mean that current implementation of drivers is not updated for it15:06
vponomaryovbswartz: drivers will try apply and some will fail15:06
bswartzhttps://review.openstack.org/#/c/40677615:06
bswartzI'm reviewing this now15:06
zhongjunbswartz: thanks ,15:08
bswartz zhongjun: ^ does this patch address the concern vponomaryov raised?15:08
*** ociuhandu has joined #openstack-manila15:09
zhongjunvponomaryov:   I will skip the tempest test if drivers will try apply and some will fail15:10
zhongjunbswartz: :)   This patch does not use the  *itertools.chain that the concern vponomaryov raised15:10
bswartzzhongjun: the better solution it to just make the drivers not fail15:10
zhongjunbswartz: cool, I'd like to, but I am not familiar with all drivers. I could call maintainer if we need15:13
bswartzzhongjun: we need to try our best15:14
zhongjunbswartz: agree15:14
bswartzin most cases all that will be needed is a is_ipv6_rule() check in the loop that looks at access rules, and skip those rules15:14
zhongjunbswartz:  Since you redesign the ensure share feature, and you are really familiars with this feature. Could take a look at this feature if you have free time https://review.openstack.org/#/c/457545/15:16
zhongjunbswartz: Added a is_ipv6_rule() check  in driver?15:17
bswartzin every driver15:18
bswartzit's the only way to gradually phase in ipv6 support15:18
bswartzwhen the driver implements ipv6 support it will remove the check15:18
zhongjunbswartz: If driver don't support ipv6 access, we will still return fail when we add a access rules to a share, even if we added a is_ipv6_rule()15:19
bswartzzhongjun: why?15:21
bswartzI don't agree15:21
bswartzlike I said before, there should be no perceivable difference between a driver that doesn't support ipv6 and a driver that does support ipv6 but doesn't have any ipv6 export locations configured15:22
bswartzin both cases, adding an ipv6 rule should be allowed15:22
bswartzthe first driver will just skip that rule in update_access() and the second driver will apply the rule but nobody will be able to tell15:22
zhongjunbswartz:  yes, adding an ipv6 rule should be allowed when a driver that doesn't support ipv615:24
zhongjunbswartz:  but run  access-allow   share_with_driver_not_support_ipv6_access   ip  rule_ipv6   ,  The rule_ipv6  state will get error.15:26
zhongjunbswartz: the first driver will just skip that rule in update_access() , but the rule state will change to error in the end.15:27
bswartzzhongjun: why is that?15:29
bswartzcan't it just go to active?15:30
zhongjunbswartz: the first driver will just skip that rule in update_access() , then the rule state will change to ?15:30
zhongjunbswartz: The active means the rule can be used for user.15:30
bswartzso what is expected to happen when you add an IPv4 rule to an IPv6-only share? do you expect an error or do you expect success?15:31
bswartzwe agreed in barcelona that it would succeed15:32
bswartzand similarly adding an ipv6 rule to an ipv4-only share should succeed15:32
bswartzthe alternative is that the user always has to know which shares support v4 and which shares support v6, which sucks15:32
zhongjunbswartz: but it is actually not succeed15:33
bswartzsucceeded just means it went into the database, and the backend did whatever it needed to do as a result of that15:33
bswartzif the backend didn't need to do anything, then that's fine15:34
bswartzthe only challenge this creates is what to do when a new export location is added which makes access rules that were previously irrelevant suddently matter15:35
zhongjunbswartz: It could be hard to explain to the user15:39
zhongjuntbarron: hi15:42
zhongjuntbarron: I saw your comment in https://review.openstack.org/#/c/465055/15:43
tbarronzhongjun: ack15:44
zhongjuntbarron: Do you mean we need to get time when we run 'df' command15:45
tbarronzhongjun: I think if we don't store timestamps with the usage numbers when we collect the info may be stale as you say.15:45
tbarronzhongjun: yes15:45
tbarronzhongjun: we should store the time the info is collected, not the time when we post it15:46
zhongjuntbarron: How about we get all info by 'df' command.   The mount_path has been removed and we can not get the usage size after a while( long loop)15:51
bswartzzhongjun: I'm in favor of this approach specifically because it should be easy for users to deal with15:53
bswartzit's harder for developers to deal with, but only in 1 specific case15:53
zhongjunbswartz: Why it cannot be easy for users to deal with if we set access rule(ipv6) state to error  when the driver don't support ipv615:58
*** xyang_ has joined #openstack-manila16:00
bswartzzhongjun: because that forces users to know which shares support ipv4 and which shares support ipv616:01
bswartzit creates a new de-facto capability that clients have to be aware of16:02
bswartzthe goal of this design was to avoid create yet more capabilities16:02
bswartzwe have enough of those on the share types16:02
bswartzeverywhere we can avoid creating new ones, we should -- and this is one place we can avoid it16:03
zhongjunWe don't need to add new capabilities since we already have ipv6_support and ipv4_support capabilities in our code16:08
bswartzuser-visible capabilities16:14
zhongjuntbarron: vkmc added the comment about add total size for the share in gather usage size feature. Do you have a strong idea about this?16:14
tbarronvkmc: ^^^^^16:15
tbarronvkmc: we already have the nominal share size from when it was created, so the question would be "why collect it again in real time"?16:19
tbarronzhongjun: vkmc still would we want to send the share size along with the usage to telemetry in the notifications?  (however we get it)?16:19
tbarronzhongjun: vkmc that sounds worth thinking about16:20
tbarronzhongjun: vkmc somebody looking at a telemetry dashboard might want to track both over time, I think that's the fundamental point vkmc is makiing there16:20
arnewiebalcktbarron: I guess the answer is related to the question what the share usage shall be used for16:21
tbarronzhongjun: and since the driver is supplying this info and it can't call into the DB for the size, maybe getting it that way would be useful.16:22
tbarronarnewiebalck: right, what are your thoughts?16:22
tbarronarnewiebalck: share usage will change over time, but share size can too.16:22
arnewiebalcktbarron: if it is for billing (as per the blueprint), you probably don’t care about the total16:22
arnewiebalcktbarron: if you want it for operators, the total is needed eventually, and here you could get it almost for free (even if you know already)16:23
arnewiebalcktbarron: for the notifications, you probably also want a backend total16:23
arnewiebalcktbarron: with a direct call, if possible, not by looping over all shares16:24
arnewiebalcktbarron: but maybe this is a different topic16:24
arnewiebalcktbarron: right16:24
zhongjuntbarron: arnewiebalck: yes,  as we written in spec,  for billing16:24
tbarronarnewiebalck: yeah, backend total should definitely be a direct call, and it is a difft topic16:24
arnewiebalcktbarron: zhongjun: for billing, actual use is probably the only metric relevant, no?16:25
tbarronzhongjun: arnewiebalck vkmc I guess if we have another use case emerge we could add share size later16:26
tbarronzhongjun: arnewiebalck vkmc it just seemed efficient to do it while we have the patient in surgery16:26
tbarronzhongjun: arnewiebalck vkmc I can imagine operator looking at time series graphs with share usages and share sizes plotted on different lines16:27
tbarronand averages of these, etc.16:27
arnewiebalcktbarron: zhongjun: total could be also be used to allow/introduce/control backend oversubscription?16:27
tbarronso I suspect someone will ask for it eventually16:27
tbarronbut zhongjun could argue back that this is "gold plating" the feature that16:28
tbarronshe signed up to do16:28
arnewiebalcktbarron: yeah, you would need these if you (as an operator) want to be in control of how much you oversubscribe a backend16:28
tbarronand she'd be right, I certainly am not doing -1 over this aspect :D16:28
arnewiebalckagreed :)16:29
arnewiebalckI was just wondering how this fits together with the ceilo work and the usage metrics there?16:29
zhongjuntbarron:  arnewiebalck:agree with not necessary add it in this feature16:30
arnewiebalckzhongjun: being just hit by a periodic task that loops on objects in the backend in another project, I was happy to see there are config options to control the share usage stats gathering :)16:31
arnewiebalcki.e. the rate and an overal enable/disable switch16:31
zhongjunarnewiebalck: Do you need total size in your platform16:31
arnewiebalckzhongjun: by “total size” you mean the sum of the usage of all shares?16:32
zhongjunarnewiebalck:  the total size of one share16:32
tbarroni think "total size" may not be the best term: nominal size of each share16:32
tbarronwhat "size" argument gives you in 'df' command16:33
arnewiebalckthe nominal size Manila (and hence I) already know, so I’d say ‘no’16:33
zhongjunarnewiebalck:  We have config options to control the share usage stats gathering16:33
arnewiebalckI think vkmc’s suggestion was to collect it as it is there for free16:34
tbarronvkmc: zhongjun arnewiebalck it's free for lvm driver where 'df' command is being used but16:34
tbarronmaybe for some backends it would be an extra call16:35
arnewiebalcktbarron: alos true, yes16:35
tbarronand that make take time16:35
arnewiebalcks/alos/also16:35
tbarronso I'm leaning to leaving it out for now16:35
vkmchey o/16:35
vkmcsorry I was on a meeting16:35
vkmcyes... my main point was that... if we can add it for free, it would be nice to have16:35
zhongjuntbarron : yes, It would be an extra call in other drviers16:35
* tbarron saw what arnewiebalck got it with in cinder with per volume periodic processing16:36
vkmcbut certainly nothing that blocks this16:36
tbarron*hit with16:36
vkmcand certainly I was thinking on lvm, not sure how that should be handled on other drivers16:36
arnewiebalckvkmc: for CephFS, I think the share size would be the quota16:37
tbarronarnewiebalck: ack16:37
tbarronbackend cephfs quota16:37
arnewiebalcktbarron: yes16:37
zhongjunvkmc: for huawei driver, it will have it's own command16:37
vkmcI see this from the operator point of view in which if you inspect data with the interest of monitoring your shares, checking only how much used it is without knowing how big your share is will lead to not actually knowing how occupied your share is16:37
tbarronnot manila quota, just in case anyone reading would be confused16:37
*** lpetrut__ has quit IRC16:38
arnewiebalckthe operator does not worry about per share fill status, I think16:38
arnewiebalckit’s more the backend fill status16:38
arnewiebalckper share would be more the user16:38
vkmcgood point16:39
tbarronthis is a good discussion: I think it leads to leaving out the nominal-size from the usage collection, at least for now16:39
vkmcall right16:40
zhongjuntbarron : +116:40
arnewiebalcktbarron: sounds good to me16:40
tbarronzhongjun: but I do think we need timestamps from the time of data collection, not just a time on the notification itself16:40
arnewiebalcktbarron: for now ;)16:40
tbarronzhongjun: arnewiebalck has thousands of shares say (he does have that many cinder volumes)16:41
zhongjuntbarron: yes, gather_time=....16:41
tbarronzhongjun: so there can be a skew16:41
tbarronover the course of the loop over shares16:42
zhongjuntbarron: "gather_at"16:42
tbarronzhongjun: ok, seems right16:42
zhongjuntbarron : ok16:42
arnewiebalckgotta run o/16:42
*** ociuhandu has quit IRC16:48
*** dustins has quit IRC17:01
*** harlowja has joined #openstack-manila17:24
openstackgerritzhongjun proposed openstack/manila master: Add share usage size tracking  https://review.openstack.org/46505517:40
openstackgerritzhongjun proposed openstack/manila master: Add share usage size tracking  https://review.openstack.org/46505517:44
*** dustins has joined #openstack-manila17:53
*** jprovazn has quit IRC17:59
*** dsariel has quit IRC18:06
*** dsariel_ has quit IRC18:07
openstackgerritBen Swartzlander proposed openstack/manila master: Allow 2 or more export IPs for LVM driver  https://review.openstack.org/44447918:09
*** deep-book-gk_ has joined #openstack-manila18:28
*** eharney has quit IRC18:30
*** deep-book-gk_ has left #openstack-manila18:30
*** eharney has joined #openstack-manila18:30
*** chlong_ has joined #openstack-manila19:03
*** chlong has quit IRC19:05
bswartztbarron gouthamr: pls re-review https://review.openstack.org/#/c/444479/19:21
*** lpetrut__ has joined #openstack-manila19:24
openstackgerritOpenStack Proposal Bot proposed openstack/manila master: Updated from global requirements  https://review.openstack.org/48455619:36
openstackgerritGoutham Pacha Ravi proposed openstack/manila master: NetApp cDOT: Fix share specs on migration  https://review.openstack.org/48493220:00
openstackgerritGoutham Pacha Ravi proposed openstack/manila master: NetApp cDOT: Add support for QoS/throughput ceilings  https://review.openstack.org/48493320:00
*** xyang_ has quit IRC20:05
openstackgerritBen Swartzlander proposed openstack/manila master: Allow 2 or more export IPs for LVM driver  https://review.openstack.org/44447920:05
*** xyang_ has joined #openstack-manila20:11
*** xyang_ has quit IRC20:13
*** lpetrut__ has quit IRC20:38
*** xyang_ has joined #openstack-manila21:07
*** xyang_ has quit IRC21:08
*** MVenesio has quit IRC21:09
*** dustins has quit IRC21:14
*** eharney has quit IRC21:40
*** jmlowe has quit IRC21:46
*** MVenesio has joined #openstack-manila21:50
*** MVenesio has quit IRC21:50
*** gouthamr has quit IRC21:53
*** gouthamr has joined #openstack-manila22:29
*** gcb has quit IRC23:02
*** ganso has quit IRC23:21
*** catintheroof has quit IRC23:25
« Monday, 2017-07-17 Index Wednesday, 2017-07-19 »

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