Это видео недоступно.
Сожалеем об этом.
How to replace dead NODE in Proxmox Cluster
HTML-код
- Опубликовано: 14 авг 2024
- How to replace NODE inside Proxmox Cluster. Surprisingly it is an easy task. In this video i will show you how to get this done in easy mode.
Proxmox Cluster setup
• Setting up Proxmox CLU...
-- Chapters
00:00 Intro
00:30 Node crashed
01:11 Node Swap process
05:02 IMPORTANT! Do not switch removed NODE ON
06:25 Add new NODE to Cluster
08:46 The End Chat
#proxmox
#proxmoxcluster
#how-to
Thank you for this video! I’ve been playing with clusters on repurposed thin clients and this explains a lot. Clusters and HA are a bit finicky on Proxmox.
At start, Cluster and HA setup was a bit "over my head" kinda thing. Now i can't live without cluster or ha. I don't know what should happen for me to go back to single proxmox node setup.
Another great video ! Thank you, much appreciated.
DANG Awesome! You finally covered something that a lot of us have struggled to figure out! Thanks MR.P 🥳
Took me some time to get all this together. I messed up so many times trying all this :)
I had to swap NODE in my main proxmox cluster - twice. If it worked for me, twice, without fail - its time to share with everybody :)
@@MRPtech Thank you for sharing this knowledge! I will save this video just in case youtube takes it down, it's pretty much gold! 🤩
However what happens if you want to re-add the node you took down with the same name again?
Adding node with same name can cause some connection issues. Each node communicates to each other and one of the ways it does - SSH. When ssh connection established at the start - each node saves all nodes names and SSH key fingerprint. When you add a new node with same name - SSH Key might conflict and you will have to fource SSH key update on all nodes. This can be done by going in each NDOE console and connecting to each node via SSH which will force SSH key update. For example. From node 1 console connect to all other nodes: ssh root@node2 / ssh root@node3.
Nice video!!
Thanks.
HI, Thanks for your video. Just one question if I am using Ceph, may I used the same process for swapping a node ?
Hi,
There are extra steps you need to do to swap NODE while using CEPH.
If node is functioning and you want to swap it, 1st you need to detach OSD from CEPH pool and "destroy" it before removing a node. This will allow CEPH to recalculate bits and move data out of a drive inside a NODE you want to remove it.
If node is dead and you want to replace it - there is a bit more to do to get this done, i will create guide video to showcase how this can be done.
Thank you so much. Waiting for this tutorial
can you name it the same as the old one or does it need to be called something different? Thanks for the content as usuall!
I always name new nodes using different names.
My guess will be - if you name new node with same hostname - you might encounter synchronization errors as new node with same name as old node will have different fingerprint.
Bad Day = Offline Node.
Really Bad Day = Removing Wrong Node from Console.
Really Really Bad Day = Removing Wrong Node from PVE folder.
Really Really Really Bad Day = Offline node switching on and some how connecting back to cluster.
In this failure scenarios, none shows the syslog flood messages in order for the users to be familiar with them (there are a couple of services that fail).
On top of that you showed the case where the node wont be active again. What about if you fixed the node and you re ready to rejoin to the cluster.
If you planing to fix a node and rejoin a cluster without reinstalling Proxmox - just don't remove it. Check why Node if OFF, maybe just power issue. If Drive is failing, clone Proxmox OS to new drive and check if its boots up without connecting to network.
@@MRPtech Thank you for the reply but didnt get the sentece
1.Check why it's node switching on ???
2.close Proxmox os to new to be drive ???
If the problem with the node is the disk then the new installation on proxmox should have the same or a different name than the previous one? So if it was proxnod1 can it stay this way or should be renamed to proxnod0 (in case none of the other nodes has the same name of course).
Sorry, i typed my reply using my phone and turns out it went off the rails with its auto-completion and auto-correction. I have edited comment to what i was trying to say in a first place.
In regards to using same node name when reinstalling proxmox. I did received that question already under this video. I will just copy/paste my answer here:
---
Adding node with same name can cause some connection issues. Each node communicates to each other and one of the ways it does - SSH. When ssh connection established at the start - each node saves all nodes names and SSH key fingerprint. When you add a new node with same name - SSH Key might conflict and you will have to fource SSH key update on all nodes. This can be done by going in each NDOE console and connecting to each node via SSH which will force SSH key update. For example. From node 1 console connect to all other nodes: ssh root@node2 / ssh root@node3.
---
But if i reinstall Proxmox on, can i give the same old name and rejoin the cluster or not?
when you replace 1 node, other 2 nodes will still have ssh key linked to that name but under different ID. They will complain that connection can't be made.
When you try to SSH from one node to the other, error message will show up informing about this connection issue and will give you command that you need to run to force SSH key ID change
Thanks MRP for the video! How can a dead node be restored from a backup created with your script?
www.youtube.com/watch?v-0V2UhPnLZKY