Hi,
My homelab is running out of ressources and instead of buying a forth server I was thinking about merging stuff down to 2 servers. Electricity in Germany is way too expensive
Right now I got 1x bare metal PVE + 2x bare metal TrueNAS servers. And I would need another PVE server to be able to run more VMs and so that I'm not totally screwed up if my only PVE server would fail so I atleast could backups of the most important VMs to another node.
Virtualizing my primary TrueNAS Server wouldn't be a big problem. I already got two Dell PERC H310 in IT mode (so basically a LSI 9211-8i HBA) so I could passthrough my 8 HDDs and 8 SSDs to a TrueNAS VM. That way I could get rid of one of my TrueNAS servers.
But the second TrueNAS server I only use for backups and it is shutdown most of the time and is only booted (over IPMI using a script) once per week to receive ZFS replications so the HDDs are totally powered down most of the time so they should live long. Now I was thining if it would be possible to virtualize that backup TrueNAS server too...
What will happen if I block the LSI drivers in PVE and PCI passthrough such a LSI 9211-8i HBA to a TrueNAS VM? If I get it right the HBAs firmware will spinup the disks at boot so they are ready to be booted from. But what happens if the drivers are then blocked once PVE has started and the autostart of the TrueNAS VM is disabled? Will the disks spindown and the heads go into parking position? Such a setup would only make sense to me if the drives are powered down as long as the TrueNAS VMs isn't started so they don't wear if the TrueNAS VM isn't started. Is such a thing possible? So what actually happens to a PCI card if it is neighter used by the host nor a guest?
Another problem I would think of is the ATX power supply. If I merge my 8 HDDs + 8 SSDs of my main TrueNAS server with the 8 SSD + 2 HDDs of my PVE server I would got 10 HDDs and 16 SSDs in one server powered by a single Consumer ATX power supply. HDDs and SSDs use 12V + 5V and I checked the datasheet of my SSDs and they use up to 4.6W on the 5V rail + up to 5.5W on the 12V rail per drive. So these 16 SSDs alone would add up to 74W on the 5V rail. IF I look at the ATC power supply rating it doesn't matter if I buy a 300W or a 1600W power supply. All ATX power supplys nowadadys will only offer 20-30A or 100-150W on the 5V rails. Only difference between a 300W and 1600W PSU is the amount of power it can handle on the 12V rail. So I'm a bit worried to not find a suitable PSU to power all those drives.
Does someone got any experiences with those two problems?
My homelab is running out of ressources and instead of buying a forth server I was thinking about merging stuff down to 2 servers. Electricity in Germany is way too expensive
Right now I got 1x bare metal PVE + 2x bare metal TrueNAS servers. And I would need another PVE server to be able to run more VMs and so that I'm not totally screwed up if my only PVE server would fail so I atleast could backups of the most important VMs to another node.
Virtualizing my primary TrueNAS Server wouldn't be a big problem. I already got two Dell PERC H310 in IT mode (so basically a LSI 9211-8i HBA) so I could passthrough my 8 HDDs and 8 SSDs to a TrueNAS VM. That way I could get rid of one of my TrueNAS servers.
But the second TrueNAS server I only use for backups and it is shutdown most of the time and is only booted (over IPMI using a script) once per week to receive ZFS replications so the HDDs are totally powered down most of the time so they should live long. Now I was thining if it would be possible to virtualize that backup TrueNAS server too...
What will happen if I block the LSI drivers in PVE and PCI passthrough such a LSI 9211-8i HBA to a TrueNAS VM? If I get it right the HBAs firmware will spinup the disks at boot so they are ready to be booted from. But what happens if the drivers are then blocked once PVE has started and the autostart of the TrueNAS VM is disabled? Will the disks spindown and the heads go into parking position? Such a setup would only make sense to me if the drives are powered down as long as the TrueNAS VMs isn't started so they don't wear if the TrueNAS VM isn't started. Is such a thing possible? So what actually happens to a PCI card if it is neighter used by the host nor a guest?
Another problem I would think of is the ATX power supply. If I merge my 8 HDDs + 8 SSDs of my main TrueNAS server with the 8 SSD + 2 HDDs of my PVE server I would got 10 HDDs and 16 SSDs in one server powered by a single Consumer ATX power supply. HDDs and SSDs use 12V + 5V and I checked the datasheet of my SSDs and they use up to 4.6W on the 5V rail + up to 5.5W on the 12V rail per drive. So these 16 SSDs alone would add up to 74W on the 5V rail. IF I look at the ATC power supply rating it doesn't matter if I buy a 300W or a 1600W power supply. All ATX power supplys nowadadys will only offer 20-30A or 100-150W on the 5V rails. Only difference between a 300W and 1600W PSU is the amount of power it can handle on the 12V rail. So I'm a bit worried to not find a suitable PSU to power all those drives.
Does someone got any experiences with those two problems?