Friday, 2017-06-09

*** jamielennox|away is now known as jamielennox00:02
*** zhurong has joined #openstack-barbican00:49
*** jamielennox is now known as jamielennox|away01:00
*** jamielennox|away is now known as jamielennox01:10
*** liujiong has joined #openstack-barbican01:22
*** hieulq has quit IRC01:23
*** hieulq has joined #openstack-barbican01:24
*** zhurong has quit IRC01:29
*** namnh has joined #openstack-barbican01:37
*** dave-mccowan has joined #openstack-barbican01:38
*** zhurong has joined #openstack-barbican02:04
*** dimtruck is now known as zz_dimtruck02:31
*** zz_dimtruck is now known as dimtruck02:52
*** dimtruck is now known as zz_dimtruck03:12
*** zz_dimtruck is now known as dimtruck03:34
*** dave-mccowan has quit IRC03:47
*** dimtruck is now known as zz_dimtruck03:54
*** zz_dimtruck is now known as dimtruck04:16
*** zhurong has quit IRC04:25
*** zhurong has joined #openstack-barbican04:34
*** dimtruck is now known as zz_dimtruck04:50
*** zz_dimtruck is now known as dimtruck05:12
*** dimtruck is now known as zz_dimtruck05:27
*** jaosorior has joined #openstack-barbican06:18
*** zz_dimtruck is now known as dimtruck06:53
*** zhurong has quit IRC06:55
*** zhurong has joined #openstack-barbican06:56
*** zhurong has quit IRC06:57
*** dimtruck is now known as zz_dimtruck07:03
*** pcaruana has joined #openstack-barbican07:05
*** zz_dimtruck is now known as dimtruck07:54
*** namnh has quit IRC07:58
*** dimtruck is now known as zz_dimtruck08:04
*** salmankhan has joined #openstack-barbican08:40
*** zz_dimtruck is now known as dimtruck08:55
*** dimtruck is now known as zz_dimtruck09:05
*** liujiong has quit IRC09:52
openstackgerritNam Nguyen Hoai proposed openstack/barbican-specs master: Specs rolling upgrade for Barbican
*** zhurong has joined #openstack-barbican09:54
*** zz_dimtruck is now known as dimtruck09:56
openstackgerritLingyong Xu proposed openstack/barbican master: Fix html_last_updated_fmt for Python3
*** namnh has joined #openstack-barbican10:03
*** dimtruck is now known as zz_dimtruck10:05
*** zhurong has quit IRC10:06
namnhjaosorior, are you free?10:13
jaosoriorsure, what's up?10:14
namnhcan I discuss with you about rolling upgrade for barbican?10:15
jaosoriorlets do that10:15
*** daidv has joined #openstack-barbican10:15
namnhjaosorior, I've just updated a specs about this.
namnhjaosorior, please take a look at it first10:16
jaosoriorok, let me do that10:17
namnhjaosorior, that would be great if the specs get your opinion about what items we need to do for rolling10:18
jaosoriorI'm reading the spec right now10:19
jaosoriorok, reading through the spec it looks fine. I mean, it's pretty high level. The real detail is in how we handle the indirection bits (which we could easily mess up)10:24
jaosoriornamnh: but it looks good! I think it does cover the overall plan. This doesn't cover yet microversions in Barbican, right?10:27
namnhyes, you are right, however Purpose of the spces is to list items what we need to do for rolling. I am considering about three items. 1 implement OVO - 2.  RPC version -3. Using trigger to solve problem during upgrade database10:29
jaosoriornamnh: right now we don't have a worker that has a clear distinction to only work with the database. Is that going to be addressed as well?10:32
namnhI have not think about that. currently there are barbican-api and barbican-worker that can interact with db. right10:34
jaosoriorboth do10:35
*** zz_dimtruck is now known as dimtruck10:35
jaosoriornamnh: I guess it's not a big deal, probably the easiest thing would be to introduce OVO for database access and just have both processes use it.10:39
jaosoriorwouldn't be as clean as we would like it to be, but at least upgrades wouldn't be that big of a hazzle, which is a huge plus10:40
*** dimtruck is now known as zz_dimtruck10:45
namnhFor OVO. in my opinion, implement OVO that depends on the architecture of project. i think OVO is used to modify information from database before sending to agents. For example, when nova-compute want to get information from database via nova-conductor then OVO on nova-condutor will modify this info as the expectation of nova-compute.10:45
namnhbut in barbican, it does not have service like nova-compute.10:46
namnhso I am considering whether we need OVO or not.10:49
jaosoriornamnh: so... The nice thing about OVO is the "make_compatible" bit. That bit we could use10:51
jaosoriorother than that, to take full advantage of it, we would need to change barbican's architecture to have a worker that's specifically handling database operations10:52
jaosoriorI don't think the RPCs are really a problem in Barbican. It's mostly database compatibility.10:53
namnhjaosorior, yead, i understood. let me expand my option about this. OVO have good method: "make_object"compatible", that is very useful in case we need to modify info from database before sending this info to agents. is that right. According to current achitecture of barbican, we dont realy need this method. are you ageed with me?10:58
namnhBut the method is useful for the future10:58
namnhso you think that we need to implement OVO for Barbican10:59
namnhis that right?10:59
jaosoriorI think it could give some benefits. However, if you think we could achieve the same thing without the "make_compatible" method, I'm actually quite open as to what we use.11:01
namnhI agree with you about OVO is usefull for the future. Let me investigate deeply about the usecase of OVO.11:07
namnhcan we jump the next topic (upgrade database)11:08
namnhjaosorior, can we jump the next topic (upgrade database)11:08
jaosoriorSo that's what I thought OVO could be useful for. Being a facade for database operations. Eventually making database upgrade easier.11:10
jaosoriorwell, not database upgrade, but upgrading barbican nodes, while using the same database.11:10
namnhjaosorior, ok, I agree with you about OVO, i means that we need to think about the strategy when the database can be changed in the future like drop column/table or add column/table.11:16
namnhCurrenlty, there are two way to solve this:11:16
namnh1. Only add database, don't allow delete or alter colume or table like nova or neutron11:19
namnh2. Using trigger like Keystone and Glance11:20
jaosoriorcan you ellaborate on option 1?11:21
namnhwhich do you prefer?11:21
namnhin neutron and nova. they have a rule from Mitaka that don't allow column or table in database. they only allow add column or table from N to (N+1) cycle11:24
jaosoriorTo be honest I prefer option 2. I rather have a more flexible and readable database, than have such a restriction; even if it involves more code.11:25
namnhthat why neutron or nova implemented OVO to modify info database to have suitable one before sending to nova-compute.11:27
namnhfor option 2. we will have 3 phases: expand, migrate, contract to upgrade database. I am sure I can implement the three one for barbican. But it have a disadvantages is that difficult to test trigger in database. that's why nova did not use trigger to upgrade database11:31
namnhin my opinion, i prefer option 211:33
jaosoriorlets do that11:33
namnhok, I will start to do this on next week. At that time, could you help me to review my patchs as main reviewer.11:36
*** zz_dimtruck is now known as dimtruck11:36
namnhjaosorior, thank you so much in advance :D11:37
jaosoriorno, thank you for this work :D this surely is very useful stuff11:37
namnhlet's jump to final topic.11:37
namnhRPC version.11:37
namnhI implemented POC for RPC version on this patch: do I need to say the overall of this iead?11:39
jaosoriorI'll figure it out, no worries11:40
namnhso that's 3 topics that I would like to discuss with you.11:42
namnhlet me sumary: 1. we will implement OVO  -- 2. Using trigger to upgrade database --  3. RPC version.11:43
jaosoriorOK, I'll review the RPC version patch as soon as I can11:44
namnhI think that's 3 items that need to implement first to support rolling upgrade. Do you have any considering?11:44
jaosoriorThe tricky part will be to test rolling upgrades11:45
*** dimtruck is now known as zz_dimtruck11:45
namnhcurrently, I don't have any idea about this. Do you have any idea?11:46
jaosoriorI'm in the same boat; I don't know if any other projects are successfully testing this. Or will we be the first if we do.11:48
*** raildo has joined #openstack-barbican11:49
namnhjaosorior, ok we will think about this later. :)11:51
namnhjaosorior, I think it is enough for today, I have to leave my office, it's so late in my country11:52
namnhjaosorior, thanks for your time to discuss with me.11:52
*** noslzzp has joined #openstack-barbican11:54
jaosoriorsure man, no biggie11:54
namnh:) see you on next week11:54
*** namnh has quit IRC11:55
*** cpuga has joined #openstack-barbican12:22
*** cpuga_ has joined #openstack-barbican12:30
*** noslzzp has quit IRC12:30
*** noslzzp has joined #openstack-barbican12:31
*** cpuga has quit IRC12:34
*** salmankhan has quit IRC12:35
*** zz_dimtruck is now known as dimtruck12:36
*** jaosorior has quit IRC12:37
*** catintheroof has joined #openstack-barbican12:40
*** dave-mccowan has joined #openstack-barbican12:43
*** catintheroof has quit IRC12:45
*** dimtruck is now known as zz_dimtruck12:46
*** catintheroof has joined #openstack-barbican12:46
*** noslzzp has quit IRC13:16
*** noslzzp has joined #openstack-barbican13:19
*** zz_dimtruck is now known as dimtruck13:37
*** chlong has joined #openstack-barbican13:52
*** dimtruck is now known as zz_dimtruck13:53
*** jaosorior has joined #openstack-barbican13:56
*** noslzzp_ has joined #openstack-barbican14:00
*** noslzzp has quit IRC14:01
*** zz_dimtruck is now known as dimtruck14:10
*** noslzzp_ has quit IRC14:17
*** salmankhan has joined #openstack-barbican14:24
-openstackstatus- NOTICE: The Gerrit service on is being restarted now to clear an issue arising from an unanticipated SSH API connection flood14:55
*** jaosorior has quit IRC15:09
*** noslzzp has joined #openstack-barbican16:13
*** pcaruana has quit IRC16:34
*** noslzzp has quit IRC17:19
*** salmankhan has quit IRC17:52
*** noslzzp has joined #openstack-barbican17:58
*** noslzzp has quit IRC19:05
openstackgerritOpenStack Proposal Bot proposed openstack/barbican master: Updated from global requirements
*** salmankhan has joined #openstack-barbican21:15
*** dimtruck is now known as zz_dimtruck21:25
*** raildo has quit IRC21:39
*** cpuga_ has quit IRC21:46
*** salmankhan has quit IRC21:56
*** chlong has quit IRC21:58
*** dave-mccowan has quit IRC22:06
*** dave-mccowan has joined #openstack-barbican22:07
*** catintheroof has quit IRC22:42
*** zz_dimtruck is now known as dimtruck23:17

Generated by 2.15.3 by Marius Gedminas - find it at!