On Fri, Nov 6, 2020 at 11:10 AM Marcin Sobczyk <msobczyk@redhat.com> wrote:
On 11/5/20 2:03 PM, Yedidyah Bar David wrote:
On Thu, Nov 5, 2020 at 2:40 PM Marcin Sobczyk <msobczyk@redhat.com> wrote:
On 11/5/20 1:22 PM, Vojtech Juranek wrote:
IMO OST should be made easy to interact with from your main development machine. TBH I didn't see much interest in running OST on developers' machines. as it's little bit complex to setup? Definitely, that's why the playbook and 'setup_for_ost.sh' was created. I hope it will be the long term solution for this problem.
Making it more easy maybe would increase number of people contributing to OST ... But that's a chicken and egg problem - who else is going to contribute to OST if not us? If we want the setup to be easier, then let's work on it.
I agree, but this requires time.
Most of the work done on OST so far was for CI. Using it by developers was a side-interest.
People are mostly using manual OST runs to verify things and that is what most of the efforts focus on. It's not that I wouldn't like OST to be more developer-friendly, I definitely would, but we need more manpower and interest for that to happen.
> I noticed that many of you run OST in a VM ending up with three layers > of VMs. > I know it works, but I got multiple reports of assertions' timeouts and > TBH I just don't > see this as a viable solution to work with OST - you need a bare metal > for that. Why?
After all, we also work on a virtualization product/project. If it's not good enough for ourselves, how do we expect others to use it? :-) I'm really cool with the engine and the hosts being VMs, but deploying engine and the hosts as VMs nested in other VM is what I think is unreasonable. I tried this approach in past two days and works fine for me (except the fast it's slow) (One of the best freudian slips I noticed recently)
Maybe I'm wrong here, but I don't think our customers run whole oVirt clusters inside VMs. There's just too much overhead with all that layers of nesting and the performance sucks.
Also, using bare-metal isn't always that easy/comfortable either, even if you have the hardware. I'm very happy with my servers. What makes working with bm hard/uncomfortable? IMHO main issue is lack of HW. I really don't want to run it directly on my dev laptop (without running it inside VM). If I ssh/mount FS/tunnel ports to/ from VM or some bare metal server really doesn't matter (assuming reasonable Lack of HW is definitely an issue, but not the only one.
When I did use BM for OST at home, my setup was something like this:
(home router) <-wifi or cable-> (laptop) <-cable-> (switch) <-cable-> BM
I configured on my laptop PXE for reinstallations. Worked quite well even if not perfect.
I worked quite hard on _not_ using my laptop as a router. I ran on it squid + squid-rpm-cache, and I think I never managed to make OST work (through a proxy). Eventually I gave up and configured the laptop as a router, and this worked.
Before that, I also tried disabling dhcp on the router and running dhcp+dns PXE on a raspberry pi on the home network, but this was problematic in other ways (I think the rpi was simply not very stable - had to power- cycle it every few months).
I can't use the router as PXE. Hmmm maybe we have a different understanding on "bare metal usage for OST". What I meant is you should use a physical server, install lago et al on it and use it to run OST as usual.
Yes, that's what I meant too.
I didn't mean using a separate bare metal for each of the engine and the hosts. That way you don't need to have your own PXE and router - it's taken care of by lago and prebuilt images of VMs.
I used these BM machines in the past for various different uses (engine/host, el7/8/fc, etc.) so at some point decided to invest in configuring PXE. If indeed I consider OST to cover my entire spectrum, it's probably enough to install from a USB stick.
I agree. I have the privilege of having separate servers to run OST. Even though that would work, I can't imagine working with OST on a daily basis on my laptop. I wonder why. Not saying I disagree.
OST eats a lot of resources and makes your machine less responsive. Given that, and the lengthy debugging cycles, with a separate server I'm able to work on 2-3 things simultaneously.
I find the biggest load on my laptop to be from browsers running web applications. I think this was ok also with my previous laptop, so assume I can restrict the browsers (with cgroups or whatever) to only use a slice (and I usually do not really care about their performance). Just didn't spend the time on trying this yet.
(Previous laptop was from 2016, 16GB RAM. Current from 2019, 32GB).
That also kinda proves my point that people are not being interested in running OST on their machines - they don't have machines they could use. I see three solutions to this: - people start pushing managers to have their own servers Not very likely. We are supposed to use VMs for that.
- we will have a machine-renting solution based on beaker (with nice, automatic provisioning for OST etc.), so we can work on bare metals +1 from me.
- we focus on the CI and live with the "launch and prey" philosophy :) If it's fast enough and fully automatic (e.g. using something like Nir's script, but perhaps with some more features), also ok for me.
Right now, this is what I use mostly. For things that are much faster to verify manually, I use manual (non-OST) setup on VMs.
connection speed to bare metal server)
I can think of reprovisioning, but that is not needed for OST usage.
CI also uses VMs for this, IIUC. Or did we move there to containers? Perhaps we should invest in making this work well inside a container. CI doesn't use VMs - it uses a mix of containers and bare metals. The solution for containers can't handle el8 and that's why we're stuck with running OST on el7 mostly (apart from the aforementioned bare metals, which use el8).
There is a 'run-ost-container.sh' script in the project. I think some people had luck using it, but I never even tried. Again, my personal opinion, as much as I find containers useful and convenient in different situations, this is not one of them - you should be using bare metal.
The "backend for OST" is a subject for a whole, new discussion. My opinion here is that we should be using oVirt as backend for OST (as in running oVirt cluster as VMs in oVirt). I'm a big fan of the dogfooding concept. I fully agree. Already talked about this with Sandro some years ago, never did anything...
This of course creates a set of new problems like "how can developers work with this", "where do we get the hosting oVirt cluster from" etc. Whooole, new discussion :) I don't think this is related to whether you use oVirt, or lago/vagrant.
If every developer has to run an oVirt engine just for running OST inside it, this indeed has some extra overhead, but if it was the only problem we'll probably still prefer it compared to re-implementing parts of it in lago.
For CI, that's a non-issue, IMO.
I'd rather meant having a fixed oVirt installation that's shared by everyone.
We do have inside the office labs such setups, but AFAIU most people do not have rights to create VMs there, only use them.
In this scenario people would create the OST VMs on that oVirt instance instead of running them locally with lago.
So what you mean is: 1. People would ask for VMs for engine/hosts 2. Would then spend some time installing software on them etc. 3. Then would tell OST to use these machines, passing credentials etc. ? What I thought you meant, and I want, is: 1. People would have rights to automatically create VMs there with their own credentials 2. Would give these credentials to OST, and it will create the VMs for them and run the tests, all fully automatically This, obviously, requires quite some investment: 1. We need to decide that it's ok 2. We need to make sure we have reasonable means to not overload the system - so that these OST VMs will often be very short-lived - e.g. either automatically decommissioned after it finishes (perhaps e.g. after 24 hours, to allow some debugging/assessment time), or have some effective agreed-upon notification to remind people to manually decommission after they are done. 3. Add the support for this in the code. Of course, we can start with (3.), and then be able to do this each on their own machines (BM or VM).
Best regards,
Regards, Marcin
> On my bare metal server OST basic run takes 30 mins to complete. This is > something one > can work with, but we can do even better. > > Thank you for your input and I hope that we can have more people > involved in OST > on a regular basis and not once-per-year hackathons. This is a complex > project, but it's > really useful. +1!
Thanks and best regards,
>> Nice. >> >> Thanks and best regards, > [1] > https://github.com/lago-project/lago/blob/7bf288ad53da3f1b86c08b3283ee9c5 > 118e7605e/lago/providers/libvirt/network.py#L162 [2] > https://github.com/oVirt/ovirt-system-tests/blob/6d5c2a0f9fb3c05afc854712 > 60065786b5fdc729/ost_utils/ost_utils/pytest/fixtures/engine.py#L105
-- Didi