installing on ESXi #8
Comments
|
You will need to enable nested virtualization on your ESXi hypervisor, which I assume can be done, but haven't done it personally. Give your snc-host 32GB of RAM, 4 or preferably 6 CPUs, and 500GB of disk. You can get away with less disk if needed. |
|
I assigned reachable address to the snc-host and edited the files so both of the bootstrap and master nodes will have IPs from the same range as well. For this reason I thought there shouldn't be necessary to create a network bridge. But DeployOkdSnc.sh script stopped giving me the following error: Should I create a network bridge in this case or edit the script to ignore bridge config (I don't know how, anyway)? |
|
The scripts that I prepared are very opinionated toward a Libvirt install on a bare metal host. You will either have to create the bridge device on your snc-host, or modify the script to build the VMs natively on ESXi. Unfortunately, I don't have an ESX setup to test with, just my bare metal NUCs. |
|
Hello; while issuing "oc apply -f", "oc adm policy" and "oc patch" commands, I got this: The oc get pods command ran successfully: htpasswd secret was also created: |
|
Hello; I restarted the host but after it booted up, even "oc get pods --all-namespaces" command which ran before, didn't run after restart, and GUI didn't work too. I tried the curl and got this: Nothing works as expected. Did I miss something? |
|
Hello; |
|
Hey, that sucks that you couldn't get it working. I don't know if the nested virtualization is causing issues or not. It looks like you got really close though. Those pods that were in a crash loop backoff state were probably the issue. If you can get back to that state, I'll try to help you get across the line. I'm in meetings all day most days, so I haven't been able to spend much time in my lab lately.. :-( |
|
I really appreciate you for helping friends despite you're steeped in the daily work. So first of all I want to say thank you :) I gave one more shot and noticed that there might be an issue in the "sshkey: ecdsa-sha1xxxxxx" line of the install-config-yaml file. I'm not sure if it makes any sense, I put the whole line inside ' ' characters (like sshkey: 'ecdsa-shaxxxxx' but got the same results as I said (bootstrap and master didn't shut down by script and I got nothing on the virsh console through the whole process), |
|
I'll take a look at that. I might have a mistake in the document. |
|
Did you get the chance to test-run the procedure on an ESXi host? I can even give you access to my lab if you interested to get info, whenever your time allows. |
|
Hey. OKD 4.7 is being released. It might work better for you. I have not had any testing time lately, so I haven't tried anything yet. I'll keep you posted. |
|
Try this again with OKD 4.7. Follow my guide, there are three oc commands that you need to run during the installation for it to complete successfully with a single node. |
Hi Sir, May i know the minimum Disks for bootstrap and master node? In your scripts, it uses 200G for each node, can i use 30G for testing? i had tried and seemed not working. |
|
I have successfully built with 50G disks. This guide is also horribly out of date. I need to refresh it for OKD 4.9. I'll try to spend a few minutes on it soon. |
thank you , please let me know the result. Still have issues with bootstrap node, find some no space errors in dmesg, but not sure whether it is caused by disk. If it is disk issues, i will create another VM for testing but i do have disk resources limitations in my notebook. |
|
How much memory are you giving the bootstrap node? It needs at least 14GB RAM. Not because it uses that much, but because it sizes the RAM disk that it expands the ostree bundle into proportionally to the RAM. If it's too small, then the ostree bundle will fill up the filesystem. |
|
It should be safe to oversubscribe your RAM if you have at least 32 GB on the box. So, try giving the bootstrap node 16GB of RAM. |
|
I only has 16G memory in my notebook, seems I can overconfigure memory for bootstrap node!
Than how about Master? Same requirements?
Thanks
Sky
Sent from my mobile
On Dec 23, 2021, at 7:22 PM, Charro Gruver ***@***.***> wrote:
It should be safe to oversubscribe your RAM if you have at least 32 GB on the box. So, try giving the bootstrap node 16GB of RAM.
—
Reply to this email directly, view it on GitHub<#8 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AVKOO3LE7NFVDWVDVCFMRY3USMA7NANCNFSM4WZRXIAQ>.
Triage notifications on the go with GitHub Mobile for iOS<https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android<https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>.
You are receiving this because you commented.Message ID: ***@***.***>
|
|
Yeah, 16GB of RAM is going to be really tight. Ideally, you need 32GB of RAM to have a useful setup. |
|
yes, seems still can cheat KVM with 16G memory even only having 12G memory in the host. |
|
bookstrip node works fine but not master node. I use virsh console to login master, always report below errors: [ 2001.156252] ignition[452]: GET error: Get "https://api-int.okd4-snc.snc.test:22623[* ] A start job is running for Ignition (fetch) (32min 59s / no limit) I also can't use ssh core@okd4-snc-master, so I guess maybe I need real "16G" memory to create master node. Anyway, many thanks for your support, I will try to install it after i got power server. |
|
just found api-int.okd4-snc.snc.test is in boostrip node, so this problem should be on bootstrip. |
|
That error message is to be expected until the bootstrap node is fully up and serving the manifests for open shift install. Once the bootstrap node is fully up, it will be listening on port 22623 to serve the API for the master node to install. The master node will keep retrying that HTTP get until it is able to retrieve its ignition file. So, I believe that your bootstrap node has not completely started up yet. You definitely need more RAM. That's why I use the little Intel NUC servers. They are perfect for this, and really portable. |
|
yes, you are right, i found that kubelet.service could not be started in bootstrap node. I also use the newest version of OKD: 4.9.0-0.okd-2021-12-12-025847 ec 24 13:15:55 okd4-snc-bootstrap.snc.test systemd[1]: kubelet.service: Scheduled restart job, restart counter is at 4. |
If I want to install the whole system on ESXi server and start to build an initial snc-host VM on it, what would be the specs of the VM (for example, disk)? If I understood right, an initial machine (snc-host) is deployed first, then okd machines (bootstrap and master/worker) are deployed as VMs on top of it. Am I right?
The text was updated successfully, but these errors were encountered: