Good evening, let me start by saying that I am using Proxmox 8.1. I have a two node cluster, where I have enabled VM replication. The cluster communicates on 1GB LAN cards. The two servers also have SFP+ ports so I would like to use fiber cables. Moving the operation of the cluster and replicas to the connection on the SFP+ ports means changing the configuration of the sluster. I think this operation is delicate, do you think I could have any problems?
Move cluster operation from 1GB LAN to 10GB SFP+ ports
- Thread starter Fabio
- Start date
You are using an out of date browser. It may not display this or other websites correctly.
You should upgrade or use an alternative browser.
You should upgrade or use an alternative browser.
I have had bad luck with SFP transceivers that provide for a copper interface. In my experience, they have a tendency to overheat and then fail intermittently. They work in principle, but in practice they are often more of a headache than what I am willing to deal with.
Fiber transceivers are a lot more forgiving from what I can tell, and you can cheaply buy pre-terminated fiber cables in a large variety of lengths. This is absolutely the way to go for connections within a single rack. A DAC cable addresses the same needs and could quite possibly be implemented over fiber. But you don't really know. And from my limited experience, DAC cables are a mixed bag. Some work fine, whereas others have intermittent problems similar to what I encountered with copper transceivers. I ended up ripping out all my DAC cables and things are much more stable now.
Things are a little less obvious for when you need to run cabling through the entire house. Fiber would be ideal here, but it is harder to deploy than copper. This is a discussion that you need to have with your electrician. But something like https://sschueller.github.io/posts/wiring-a-home-with-fiber/ is a good starting point.
Fiber transceivers are a lot more forgiving from what I can tell, and you can cheaply buy pre-terminated fiber cables in a large variety of lengths. This is absolutely the way to go for connections within a single rack. A DAC cable addresses the same needs and could quite possibly be implemented over fiber. But you don't really know. And from my limited experience, DAC cables are a mixed bag. Some work fine, whereas others have intermittent problems similar to what I encountered with copper transceivers. I ended up ripping out all my DAC cables and things are much more stable now.
Things are a little less obvious for when you need to run cabling through the entire house. Fiber would be ideal here, but it is harder to deploy than copper. This is a discussion that you need to have with your electrician. But something like https://sschueller.github.io/posts/wiring-a-home-with-fiber/ is a good starting point.
I should use these cables: https://icecat.biz/it/p/dell/470-aavh/fibre+optic+cables-5397063816781-470-aavh-25589223.html.
I can confirm that, I would definitely not recommend an RJ45 adapter.I have had bad luck with SFP transceivers that provide for a copper interface. In my experience, they have a tendency to overheat and then fail intermittently. They work in principle, but in practice they are often more of a headache than what I am willing to deal with.
I only buy cables from Flexoptix, where I have a Flexbox and can brand the transceivers the way I need them. I've only had good experiences with the cables so far. Basically, if you buy cheap, you buy twice.And from my limited experience, DAC cables are a mixed bag. Some work fine, whereas others have intermittent problems similar to what I encountered with copper transceivers. I ended up ripping out all my DAC cables and things are much more stable now.
For the cabling in the rack I would always prefer DAC, everything that goes out should be normal SM or MM.
If it's also the Dell network cards, that shouldn't cause any problems. You have to be careful with other manufacturers or brands.I should use these cables: https://icecat.biz/it/p/dell/470-aavh/fibre+optic+cables-5397063816781-470-aavh-25589223.html.
I had the same idea. For redundancy reasons I tried to move the traffic on vmbr0 from 1x 1GbE (eno1) to a 2x 10GbE (bond0 alb), but afterwards the host could not reach its default gateway anymore, though other devices on the subnet could be pinged, except for the the other proxmoxes. In the end, I left the bond for the VMs and the 1GbE for the cluster.
I'm using RJ45 instead of fibre, but no SFP+ plugs.
I'm using RJ45 instead of fibre, but no SFP+ plugs.