[10:45] <gsimondo1> Got a 3 machine juju cloud that is currently stuck because it can't resolve a relation between influxdb and telegraf. This puts influxdb in error state. Resolving its state just puts it back in error state. The consequence is that I can't remove model, controller, machine, or anything that's above this unit/relation in stack.
[10:45] <gsimondo1> Any advice on how to handle these situations?
[11:03] <kjackal> hi gsimondo1, first file a bug with the proper section of juju debug-log showing the error. Then you can remove the machine where the failing application is deployed and that should unblock all the remove operations
[11:04] <kjackal> let me check if there is a --force flag
[11:04] <kjackal> yes, juju remove-machine <id> --force
[13:22] <gsimondo1> kjackal: Just saw your message. I figured that out a couple of hours ago but that kills the machine. So I take it that this is a bug that's worth reporting on github?
[13:24] <kjackal> gsimondo1: if you want to keep the machine you can try juju resolved with --no-retry flag
[13:26] <gsimondo1> kjackal: doesn't help. it wakes up and reexecutes the failing relation. can't get rid of the relation.
[13:26] <gsimondo1> 2018-02-12 13:25:55 DEBUG query-relation-joined TypeError: configure() missing 2 required positional arguments: 'username' and 'password'
[13:26] <gsimondo1> 2018-02-12 13:25:55 ERROR juju.worker.uniter.operation runhook.go:113 hook "query-relation-joined" failed: exit status 1
[13:30] <kjackal> gsimondo1: this is on the telegraf or the influx db side?
[13:30] <kjackal> influxdb, sorry
[13:30] <gsimondo1> root@kubernetes-2 ~ # tail -f /var/log/juju/unit-influxdb-0.log
[13:32] <kjackal> gsimondo1: how did you deploy this charm?
[13:32] <kjackal> I am looking for the revision
[13:32] <kjackal> is it thisone: https://jujucharms.com/influxdb/13
[13:33] <gsimondo1> kjackal: ehalilov@kubernetes-0 ~ $ juju deploy cs:~influxdb-charmers/influxdb
[13:33] <gsimondo1> before that juju deploy telegraf
[13:35] <gsimondo1> kjackal: the problem is of course that I was experimenting with relations that are not offered when you do 'juju add-relation telegraf influxdb'
[13:36] <gsimondo1> kjackal: with something like 'juju remove-relation telegraf:juju-info influxdb:juju-info'
[13:36] <gsimondo1> *add-relation, not remove
[13:38] <kjackal> gsimondo1: the error you got says that configure() missing 2 required positional arguments: 'username' and 'password', this method is here: https://github.com/ChrisMacNaughton/interface-influxdb-api/blob/master/provides.py#L27  and the charm is calling this method here: https://git.launchpad.net/influxdb-charm/tree/reactive/influxdb.py#n261
[13:38] <kjackal> so... this is a bug on the influxdb charm if I understand this correctly
[13:39] <kjackal> gsimondo1: will you be able to open a ticket here: https://launchpad.net/influxdb-charm ?
[13:41] <gsimondo1> kjackal: OK, I'll test a couple of other things and then open a ticket. I also have some other things failing in a similar fashion (relation related)
[13:42] <kjackal> now, to get you unblocked.... how confident are you in your scripting skills? If you juju ssh in influxdb you can find the charm source under /var/lib/juju/unit-influxdb-0(probably)/charm
[13:42] <gsimondo1> kjackal: correct me if I'm wrong but the main issue here is that I can't get rid of the failing relation... and it has effect on things like model, controller, machine, etc. Basically putting everything in state of paralysis and this is a bug that should be fixed on the level of juju or I'm missing something?
[13:43] <gsimondo1> kjackal: OK let me take a look at that file. I'm software engineer turned devops so I can code.
[13:43] <kjackal> awesome
[13:44] <kjackal> so if you go to the reactive/influxdb.py and do something like an early "return" you will not be hitting this road block
[13:48] <kjackal> gsimondo1: your issue is already reported: https://bugs.launchpad.net/influxdb-charm/+bug/1723334
[13:48] <mup> Bug #1723334: influxdb-api relation breaks when related to telegraf charm <InfluxDB Charm:New> <https://launchpad.net/bugs/1723334>
[13:49] <gsimondo1> mup: Yep
[13:49] <mup> gsimondo1: I apologize, but I'm pretty strict about only responding to known commands.
[13:51] <kjackal> mup: help
[13:51] <mup> kjackal: Run "help <cmdname>" for details on: bug, contrib, echo, help, infer, issue, login, poke, register, run, sendraw, sms
[13:51] <zeestrat> mup: help help
[13:51] <mup> zeestrat: help [<cmdname>] — Displays available commands or details for a specific command.
[13:52] <kjackal> mup: help sms
[13:52] <mup> kjackal: sms <nick> <message ...> — Sends an SMS message.
[13:52] <mup> kjackal: The configured LDAP directory is queried for a person with the provided IRC nick ("mozillaNickname") and a phone ("mobile") in international format (+NN...). The message sender must also be registered in the LDAP directory with the IRC nick in use.
[13:53] <zeestrat> What a helpful fellow
[14:00] <gsimondo1> kjackal: commenting the problematic line works but for sure not a solution to the problem
[14:13] <kjackal> yes, this has to be fixxed upstream in the charm code
[14:25] <gsimondo1> kjackal: I don't know people who use canonical k8s yet but it seems to me that it would be rather painful without custom charms
[14:26] <kjackal> gsimondo1: are you deploying k8s?
[14:27] <kjackal> I am with the team working on packaging k8s, can you elaborate a bit on the issues you faced?
[14:28] <kjackal> gsimondo1: we have a set of add-on charms that do the monitoring and logging but they are not based on influxdb, we use prometheus
[14:29] <gsimondo1> kjackal: creating a proof of concept but firstly trying to break juju as much as possible. the out of the box k8s "production" bundle provided installs fine but we would like to use components that do not have charms already or don't have them for xenial
[14:29] <gsimondo1> kjackal: I want prometheus but I want to have influxdb for durable storage
[14:29] <gsimondo1> kjackal: currently I use remote read/write that they offer for our current solution outside of k8s
[14:30] <kjackal> would you be able to make a case for influxdb addition here: https://github.com/juju-solutions/bundle-canonical-kubernetes/issues
[14:31] <kjackal> gsimondo1: this is the place where we gather all requests and bugs on canonical k8s
[14:33] <gsimondo1> kjackal: the only problem is that remote read/write for prometheus and influxdb integration seems to be something that they want to solve by providing an ability to add influxdb as default storage for prometheus
[14:33] <gsimondo1> kjackal: here's paul dix on that https://www.youtube.com/watch?v=BZkHlhautGk
[14:34] <gsimondo1> kjackal: so building something for this remote read/write solution that may be superseded by a more superior solution that's part of prometheus code seems like a waste of time
[14:34] <gsimondo1> kjackal: but I can create an issue if nothing then for discussion about this
[14:36] <kjackal> that would be nice, thank you
[14:37] <kjackal> gsimondo1: We keep kubernetes bundle addons here: https://github.com/conjure-up/spells/tree/master/canonical-kubernetes/addons
[14:38] <kjackal> i guess graylog and prometheus are the most interesting
[14:40] <gsimondo1> kjackal: so prometheus one is pretty much what I'm trying to test currently, just with influxdb in the mix for storage.
[14:40] <gsimondo1> kjackal: outside of the bundle you provide of course
[14:42] <gsimondo1> kjackal: the most dangerous thing I've encountered so far is the bug in these upstream charms that led to an impasse in my testing environment where I can't do anything due to some relation hook failing. my conclusion from that is that one has to test all the edge cases of charms used if they don't want to experience such paralyzing events in production systems
[14:43] <gsimondo1> kjackal: again, not sure if juju should provide some kind of mechanism to exit the deadlock as you seem to suggest that charms should just work, if they don't then hot fix the code
[15:56] <kwmonroe> gsimondo1: one thing you might be hitting is that events can queue up on the failing unit, so you'd have to run --no-retry multiple times.  for example, influx may fail on the query relation, but then have a peer or status event queue up behind that.   if the root cause of the error affects all those events, you'd have to run "juju resolved --no-retry influxdb/0" multiple times to get the unit back to a responsive state.
[15:57] <kwmonroe> you could watch status with "juju debug-log -i influxdb/0 --tail" in one window while you're doing the --no-retry.  you should see juju progress past one hook, then potentially move to another that you have to pass with --no-retry, etc etc.
[15:59] <gsimondo1> kwmonroe: useful, thanks for that. I'll test if that resolves the issue. So far I've been mitigating this by manually editing code to get out of the deadlock
[16:02] <kwmonroe> gsimondo1: fyi, this feels like bug 1741506.  if we had a --force flag for juju remove-*, it could do the --no-retries automatically until the removal could be completed.
[16:02] <mup> Bug #1741506: Removing a unit or application doesn't resolve errors <cli> <usability> <juju:Triaged> <https://launchpad.net/bugs/1741506>
[16:05] <gsimondo1> kwmonroe: seems like it. also, from the point of view of UI, what you suggest is important because prompting people should be optional IMO
[16:06] <kwmonroe> +1
[17:01] <rick_h> bdx: howdy
[19:29] <bobeo> rick_h: bdx So I deploye two instances of owncloud into juju,and I also deployed two instances of postgresql, and the databases have synced as expected, and are replicating as expected, however, owncloud did not, even though owncloud uses postgresql for user accounts and for data storage. Can anyone explain, or help me understand something that I missed?
[19:43] <kwmonroe> bobeo: a quick look at the charm source (https://github.com/omnivector-solutions/layer-owncloud/blob/master/metadata.yaml) doesn't show any peer relations.  that means multiple owncloud instances won't do anything special when another comes along.  if you've deployed 2 of those, i'd bet that they both have their own db created and are working as 2 independent ownclouds.
[19:46] <bobeo> kwmonroe: But it does seem that the postgresql that I deployed does have it. postgresql/0*  active    idle   4/lxd/6  10.0.0.42       5432/tcp  Live master (9.5.11) postgresql/1   active    idle   5/lxd/2  10.0.0.80       5432/tcp  Live secondary (9.5.11)
[19:46] <bobeo> kwmonroe: The thing that confuses me, is that shouldnt the postgresql database be storing all of the configurations data, IE the databa OOOOO! if the webserver doesnt point to the data, it doesnt matter that its replicated!
[19:47] <bobeo> kwmonroe: so what would you recommend I do to "make it" do the thing? Its so close to perfect for me. What are my options?
[19:48] <kwmonroe> bobeo: what thing are you trying to do?  make it so that 2 owncloud units use the same database?
[19:48] <bobeo> kwmonroe: as for owncloud to the db, I added the relation of postgresql:db owncloud:postgresql
[19:48] <kwmonroe> right -- postgres does have a peer relation, so it will replicate amongst the cluster of postgres units.
[19:48] <bobeo> kwmonroe: Yes, and that the data replicates in the database between the database clusters, specifically so that 2 servers have the data in two databases, so that if one physical server dies, services remain available.
[19:49] <bobeo> kwmonroe: The idea is point haproxy forwarders to the local haproxy services, which point at the owncloud servers that share the same postgresql clusters
[19:49] <bobeo> high availablity from top to bottom. I lost a django project today. Never again.
[19:50] <bobeo> I lost almost a month worth of code progress, damn near cried, and quite literally might have bitten off something's head had anything live been within range at the moment of realization
[19:51] <kwmonroe> ouch :(
[19:52] <bobeo> kwmonroe: yea, especially because im new to django. The idea is build cool apps that solve real problems using django, decorate it using bootstrap and reactive. I was so close. less than 30 minutes away, and I would have finished my first successful django project.
[19:54] <bobeo> kwmonroe: so im hoping this idea, combined with a gitlab to go with it, will prevent that from ever happening again. Plus itll allow me to share the code with others for peer review, as well as external input/assistance.
[19:55] <kwmonroe> bobeo: since the postgres peering work is already done, the next piece would be to make owncloud aware of its peers.  i dunno what that entails, so you'd have to check the OC docs for whatever is needed to run a cluster of owncloud apps.  it may be as simple as just ensuring they all use the same db creds.
[19:56] <kwmonroe> bobeo: i think the harder part will be for the object storage.  i'm assuming OC uses some actual filesystem to store objects and that they're not just blobs in a database.
[19:56] <kwmonroe> so for that, you'd need to have a common NFS share, or s3 storage, or something so all the OC units could access the same backend data.
[19:57] <bobeo> kwmonroe: I was under the impression since it didnt include using NFS that it was storing as BLOB in the postgresql db. How do I verify if thats the case? Ive never run into not knowing how the "data" is stored before?
[19:58] <kwmonroe> bobeo: i see that bdx has storage on his todo list -- https://github.com/omnivector-solutions/layer-owncloud/blob/master/README.md#todo -- maybe sync with him to find out if there is any work in progress to help with.
[19:59] <bobeo> kwmonroe: ahhh! ill do that. Im hoping to get this thing running sooner, rather than later. My soul still burns. My poor django baby. I dont see how developers carry on. It feels like I lost a piece of my very being.
[20:02] <kwmonroe> bobeo: if you have an OC deployment handy, upload a relatively big file (100ish megs) and see where it goes on the unit.  i'm guessing /var/www/owncloud/data (which is the location that would need to be shared amongst your OC peers).
[20:02] <kwmonroe> if it does jam it into postgres, you'll probably see /var/lib/postgres grow by 100ish megs
[20:06] <kwmonroe> bobeo: looks like you can setup external storage through the GUI; https://doc.owncloud.org/server/10.0/admin_manual/configuration/files/external_storage_configuration_gui.html -- so if you get multiple OC units talking to the same database, then go in to each UI and configure external storage, that might get you where you want to be.  then you'd hopefully find a programatic way to do that config so the charm could do it for
[20:06] <kwmonroe>  you next time.
[20:12] <rick_h> kwmonroe: bobeo yea, you need the OC charm to elect a leader, pass it's pgsql connection to the non-leaders, and enable the leader sending object storage creds to the others as well
[20:12] <rick_h> kwmonroe: bobeo the idea being that if a non-leader dies no biggie and if the leader goes down a new one will be elected and already have the details it needs
[20:13] <rick_h> kwmonroe: bobeo it's a bit much to bite off as your first bits of charming and maybe you can send a note to the list and get some examples. All the ones I can think of would be openstack and rather large chunks to process
[20:17] <bobeo> rick_h: That sounds like serious "OOoh dis gonna hurt" content to try to take on. I guess in the interum simply wait until my skill and comfort level builds? And use Git instead? Its mostly the code Im concerned about. Is that a feature gitlab supports?
[21:28] <kwmonroe> bobeo: the leadership bits rick_h was referring to would need to be implemented in the owncloud charm.  the hard part of leadership is already handled by juju and layer-leadership.  the last mile is that piece in the OC charm that sets/gets config to ensure all OC units have the same info.
[21:29] <kwmonroe> i'm not quite sure what you meant with "feature gitlab supports", but again, this would be up to the owncloud charm to coordinate the config amonst its peers.
[21:31] <yosefrow_> anyone with knowledge of juju/openstack hear of a kolla juju bundle for openstack
[21:31] <kwmonroe> the docs cover a bit about juju's leadership capabilities (https://jujucharms.com/docs/stable/developer-leadership) and the layer that owncloud would need to include has a bit more about how a reactive charm works with leadership (https://git.launchpad.net/layer-leadership/tree/README.md)
[21:32] <kwmonroe> ^^ that was for bobeo, not you yosefrow_ :)
[21:32] <yosefrow_> I had an conversation yesterday trying to convince someone to switch to juju for openstack, and they insisted that Kolla is the future for openstack deployments
[21:32] <yosefrow_> i didnt really have a comeback
[21:33] <yosefrow_> pretty much the conversation came down to why Kolla (openstack with dockers) is better than juju openstack bundle (which uses lxd)
[21:34] <kwmonroe> yosefrow_: i don't have OS/Kolla exp, but typically when it comes down to docker vs lxd, it's good to consider what type of container is best for you.  docker (process) vs lxc (system) containers.
[21:35] <kwmonroe> i like to ssh into things and poke around.  having /var/log/* and "normal" systemd / init functions make system containers easier for me to diagnose/debug.  ymmv of course.
[21:35] <yosefrow_> kwmonroe, their argument was basically that lxd is dying and therefore not nearly as well supported as docker
[21:35] <bobeo> yosefrow_: I would heavily disagree with them. Nothing is stopping you from installing magnum, which will allow you to use docker containers with openstack. You dont want docker containers for infrastructure, you want them for containers for projects. System containers allow you to push process containers, or containers in containers. You cant as easily do that with lxc in docker, but very easily with docker in lxc. VM in a VM arguement.
[21:36] <bobeo> a type 1 into a type 2, but you can a type 2 into a type 1, which I have also done, and currently do.
[21:37] <kwmonroe> did i miss a netcraft report?  yosefrow_, i'd love to see the "lxd is dying" source.
[21:37] <yosefrow_> kwmonroe, if not dying, than not very popular
[21:37] <yosefrow_> thats pretty much what they said
[21:37] <yosefrow_> but i had no statistics
[21:38] <bobeo> kwmonroe: No, you didnt. I speak with the netcraft people all the time. LXD is doing better than fine these days. Docker is overfluffed peacock fluffing its feathers again as always.
[21:38] <yosefrow_> bobeo, their argument for container usage is that eventually they say openstack services will move to kubernetes because it allows seamless and 100% reliable deployment with pod failover features. Therefore, docker is the path forward.
[21:39] <yosefrow_> Is there a solid basis to say that this is not hte case, and that system services should not be provided in docker containers?
[21:40] <bobeo> yosefrow_: They are wrong. That feature already exists with Nova_Evactuate, which is also possible with Magnum. That feature has existed for years for Openstack. Not only are they full of it, they are also severely late to the game. I know kelsey personally, and he will tell you the same, as he has always told people. Kube has its purpose, so does docker, so does KVM, so does LXD.
[21:41] <bobeo> yosefrow_: Yes. The first would be my personal experience, when I had to rebuild my entire openstack environment because 1 container crapped out. Unfortunately, it was my Keystone container, the only one you cant recover from if you dont have a proper HA deployment available.
[21:42] <yosefrow_> bobeo, so the main issue with system containers is their tendency to fail, and failing to take into account that many system services cannot follow the recovery model, but must simply never go down ?
[21:42] <bobeo> yosefrow_: The second would be any number of members from the kubenetes development team. They will directly tell you kube is designed as a devops tool, not as  an infrastructure tool. Its not designed with the robustness to maintain a high availability, high load, feature rich environment. Containers, especially docker, has been service driven in design. its designed to provide one specific purpose, and perform it well. Infrastrucuture 
[21:42] <yosefrow_> Basically that the fact that kubernetes services can recover quickly is irrelevant because the services cannot afford to fail in the first place?
[21:43] <bobeo> continues to require, a robust list of capabilities, and adapt on the fly requirements, which containers arent designed to do. They are designed to be depricated, not maintained.
[21:44] <bobeo> yosefrow_: Service containers historically die, and are built around a model that thats acceptable. System containers are build around the idea of "available at all costs". One is designed for HA Enviroments to cover their weaknesses, one is designed to be the HA environment.
[21:44] <yosefrow_> bobeo beautifully put. I like this distinction.
[21:44] <bobeo> yosefrow_: also, LXC containers have a VERY LONG history of success, docker containers do not. They are very new comparatively, and people historically need to knwo things will work.
[21:45] <yosefrow_> bobeo, do you have a link I can show the next guy who questions the viability or survivability of LXD as a container platform?
[21:46] <bobeo> yosefrow_: Honestly, I would put docker inside of LXC containers, and benefit from both. Use kubernetes to manage the docker containers inside of LXC. Enjoy the benefits of LXC system performance and density improvements over KVM, and also enjoy the centralized management performance and mobility of Docker/Kubernetes. That is our current plan with our project. To migrate to that model.
[21:47] <bobeo> Sure. Check out LXD vs KVM first though. Source is Canonical. Vancouver Event. It gives a good breakdown of what KVM is, and what LXC containers are. It gives a great insight into the performance difference as well.
[21:47] <bobeo> yosefrow_: allows the other side to understand why lxc containers in the first place, and then see why LXC is very different from Docker containers.
[21:47] <yosefrow_> bobeo, my main question in this context is whether or not the future of openstack services will be docker driven. From what you've said I've gathered that even if the capability for OS services via docker exists, its a bad idea because OS is too sensitive to change to rely on the Cattle Philosophy.
[21:47] <bobeo> yosefrow_: a good way to structure the environment is think about it like Hypervisors vs Baremetal.
[21:48] <bobeo> yosefrow_: Yes, that is correct. Docker isnt designed to support drastic OS shifts. its why the current issues with docker containers is patch management. The containers die when you patch many times, whereas LXC dont give a sh*t. Patch em 50 times in an hour, LXC is Honey Badger, LXC dont care.
[21:51] <yosefrow_> bobeo, so you are aware of the challenges that projects like magnum/kolla are facing, and in your opinion they are simply barking up the wrong tree?
[21:52] <bobeo> yosefrow_: Yes. Think about it from this perspective. Instead of looking at it from a Me vs. You Perspective, if we work together to Co-Exist, we are both able to play to our strengths and weaknesses better, rather than trying to focus on being better generalists in a moot point to replace each other that will surely fail.
[21:53] <yosefrow_> bobeo, my interests are completely apolitical when it comes to solutions
[21:53] <yosefrow_> This is the line that sticks out for me: <bobeo> yosefrow_: Service containers historically die, and are built around a model that thats acceptable. System containers are build around the idea of "available at all costs". One is designed for HA Enviroments to cover their weaknesses, one is designed to be the HA environment
[21:53] <bobeo> yosefrow_: Open source devs, and projects need to realize co-existance is the best path to highest performance yield. I use a wide variety of projects that do the same thing to achieve a task, and pour my efforts and focus into interoperability, which is the easiest of dev tasks.
[21:53] <yosefrow_> bobeo, my job is to analyze, distill, and understand the spirit of things
[21:54] <yosefrow_> bobeo, I can sympathize and understand the frustration with entities that refuse to coexist as a symptom of feeling superior
[21:55] <bobeo> yosefrow_: Thats exactly the issue. Every project has this crazy idea that it needs to be "superior" whatever that means. Open source isnt the Olympics, you dont get a Gold Medal or a Good Job Cookie for being the best.
[21:56] <yosefrow_> bobeo, on the other hand, a good amount of ego feeds innovation
[21:57] <bobeo> yosefrow_: To be honest, there is no "best", only "best for the job at hand". For Us, that means using MySQL, MongoDB, and PostgreSQL at the same time. Each better at something than the other, using HAProxy, NGinx, AND Apache to get the jobs done.
[21:57] <bobeo> yosefrow_: Not really. Historically Ego has been toxic to innovation.
[21:57] <bobeo> yosefrow_: If nothing else, ego is what drives your teams apart, and pushes them to create their own project, just to spite yours.
[21:57] <yosefrow_> bobeo, I mean innovation requires a healthy amount of pride in what you do
[21:58] <yosefrow_> if you dont believe in the solution you are building, why build it
[21:58] <bobeo> yosefrow_: Its not in the best interest of the community. Yes, its excellent to take pride in what you do, but not to be proudful. Pride is the greatest path to self destruction.
[21:58] <yosefrow_> I'm not saying to ignore the next guy, just to have enough pride to drive your passion forward
[21:59] <bobeo> yosefrow_: absolutely dont ignore them, call them, hang out with them, share beers together. When projects merge, that when you see amazing things built. Almost all of our greatest OS tech came from projects merging. When great minds collide, miracles happen. Let passion be your guide, and you will find that you will fly faster than any rocket ever could.
[21:59] <yosefrow_> If in fact someone is blinding themselves because of pride, that's of course destructive behavior
[22:00] <yosefrow_> but if pride is driving someone forward causing them to believe they can do something better because they know they can, than I think its a good thing
[22:00] <bobeo> yosefrow_: Yes, but the issue is pride can be bruised and beaten, passion cannot.
[22:01] <yosefrow_> ok, I tried to play the devil's out advocate. I'm out of ammo xD
[22:02] <yosefrow_> bobeo, well I think the shift towards a more cooperative community is happening already. But there are still pockets of resistance, where some people refuse to cooperate because of their own self importance.
[22:03] <yosefrow_> bobeo, We will get there though. I'm sure
[22:03] <bobeo> yosefrow_: Lol! I learned the hard way. My pride got so large it became a sun, creating its own gravity well, crushing everything I valued, blinding me, and pushing a lot of people I cared about away. I learned that life lesson the hard way, hopefully you wont have to. And yes, there will be hold outs. And they will burn, just like Rohan.
[22:04] <yosefrow_> bobeo, been there and done that. I've been through the fires that lead to self awareness. I can't say I've completely discarded my pride. But at least I've become aware of it.
[22:05] <yosefrow_> Either way, I agree with what you said earlier, that both docker and lxd have an important role to play in the future of openstack and cloud computing in general
[22:05] <yosefrow_> I will take your advice with me to my next conversation. Thanks for all the tips :)