User talk:AKosiaris (WMF)/Benchmarking kubernetes apps
First Run Through Notes
I am using a Mac and ab might already be present on the machine. I didn't have to install anything.
- Ah, that's nice to hear. We will anyway need to differentiate that part of the guide a bit, thanks for the Mac OS input Alexandros Kosiaris (talk)
When trying to open heapster for the first time, it says the heapster annotation in a service is necessary, but this doesn't appear to be true (maybe we should note this in the guide?). The heapster page opened in my browser the second time I tried `minikube addons open heapster`, and I can see metrics for all my pods.
- Sounds like a race. The annotation probably happened asynchronously in the background by some heapster component. But yes we should probably add a note about it. Alexandros Kosiaris (talk)
However, I was confused by the directions to wait for it to start recording. In my case, I actually needed to click the 4 squares button on the top of the toolbar and navigate to pods, where I could choose my namespace and pod I wanted to look at in order to see the graphs. (I've never used this tool before so it might be obvious to someone who has)
- Hm, I had a different experience. I had to wait a couple of mins for data to actually start showing up in the graphs. I distinctly remember having to do what you did a couple of times before I got to see data. That probably is explained by the different speed the various container images are being fetched over the network. As far as navigation in app goes (it's a just a grafana after all) we can add a note about choosing the "pods" dashboard with parameter "default" for the namespace and let the user figure out the pod name themselves (it should be trivial at this point) Alexandros Kosiaris (talk)
Should this have said blubberoid instead of graphoid?: `curl https://docker-registry.wikimedia.org/v2/wikimedia/mediawiki-services-graphoid/tags/list | jq '.'` I don't have jq and this didn't work for me but I didn't think it was necessary as long as I had the right image tag in the following command.
- Yeah typo. It should have been https://docker-registry.wikimedia.org/v2/wikimedia/blubber/tags/list. Funnily enough the repo is named blubber, not blubberoid hence the diff there. But this is not expected to be the norm. As far as jq goes, it will not even be required when we start tagging images with latest (and the pipeline informs the user with the tag) Alexandros Kosiaris (talk)
When installing blubberoid with helm, I needed to be in the charts directory within deployment-charts for the provided command to work.
- Indeed, will fix. Alexandros Kosiaris (talk)
If readers are copying and pasting the commands, they might not notice that "lala" should be replaced with the name of their datafile when copying the concurrency for loop
- Indeed, I 'll make it more clear Alexandros Kosiaris (talk)
Thanks for all the input. It's been very helpful Alexandros Kosiaris (talk) 09:07, 13 December 2018 (UTC)
helm init
"init" doesn't exist in v3 of helm which is what I get with the Debian package installation instructions. Should this document clarify that helm v2 should be used or should we drop the "init" step? Kosta Harlan (talk) 10:24, 4 November 2020 (UTC)
heapster
When I try to enable heapster, I get:
❯ minikube addons enable heapster ⌛ enable metrics-server addon instead of heapster addon because heapster is deprecated 🌟 The 'metrics-server' addon is enabled
And when I try to open metrics-server, I get:
❯ minikube addons open metrics-server |-------------|----------------|-------------|--------------| | NAMESPACE | NAME | TARGET PORT | URL | |-------------|----------------|-------------|--------------| | kube-system | metrics-server | | No node port | |-------------|----------------|-------------|--------------| 😿 service kube-system/metrics-server has no node port