Hi,
Server Invironment:
Proxmox Cluster Version: 5.4.13 (diff. previous Versions tested)
7x Computing Nodes with Boot Disk SSD (normal SATA or NVMe)
3x Server Xeon 2640v3 (Dell) 1x 1-Gbit Network or 2x 10Gbit Network
4x Server AMD Epyc (1x7551,3x7401P) 10Gbit Network
2x Data Nodes (NFS 3, Debian) as Shared Storage for all Computing Nodes
each Data Node 8x 1TB GB SSD (Raid5 Data) and 8x 2TB SSD (Raid5)
Cisco 10GBit SG350XG-2F10 12-Port 10G
Problem:
If we migrate a virtual Server node online (!) e.g. Debian System from an Intel Node to an Epyc Node the Server crashes afer approx. 10min to 2 Hours with the attached screenshot failure. If we reboot the Server (on/off or reset) the System is stable and runs without failure. Tested with default KVM (i440fx) and alternative q35 Emu. (Seabios)
If we do the same with offline Migration (Server off) there is no failure.
If me migrate Server online from an AMD Epyc to AMD Epyc or Intel to Intel there is no Problem.
Tested with MTU1500/9000. No Network Problems found. No Log Problems found in Proxmox (graylog search). Diff. Memory (with Baloon or not), VirIO SCSI, Bridged Networking (no VLAN), diff. CPU (1 or multi CPU). Bios are updated (AMD are dif. Supermicro Boards)
Any Idea? I think it seems there is a differenz between Intel/AMD Host CPU with the default KVM Emu or a timing Problem.
Thanks and best regards
John Celtar
Server Invironment:
Proxmox Cluster Version: 5.4.13 (diff. previous Versions tested)
7x Computing Nodes with Boot Disk SSD (normal SATA or NVMe)
3x Server Xeon 2640v3 (Dell) 1x 1-Gbit Network or 2x 10Gbit Network
4x Server AMD Epyc (1x7551,3x7401P) 10Gbit Network
2x Data Nodes (NFS 3, Debian) as Shared Storage for all Computing Nodes
each Data Node 8x 1TB GB SSD (Raid5 Data) and 8x 2TB SSD (Raid5)
Cisco 10GBit SG350XG-2F10 12-Port 10G
Problem:
If we migrate a virtual Server node online (!) e.g. Debian System from an Intel Node to an Epyc Node the Server crashes afer approx. 10min to 2 Hours with the attached screenshot failure. If we reboot the Server (on/off or reset) the System is stable and runs without failure. Tested with default KVM (i440fx) and alternative q35 Emu. (Seabios)
If we do the same with offline Migration (Server off) there is no failure.
If me migrate Server online from an AMD Epyc to AMD Epyc or Intel to Intel there is no Problem.
Tested with MTU1500/9000. No Network Problems found. No Log Problems found in Proxmox (graylog search). Diff. Memory (with Baloon or not), VirIO SCSI, Bridged Networking (no VLAN), diff. CPU (1 or multi CPU). Bios are updated (AMD are dif. Supermicro Boards)
Any Idea? I think it seems there is a differenz between Intel/AMD Host CPU with the default KVM Emu or a timing Problem.
Thanks and best regards
John Celtar
