Proxmox in Proxmox Matrushka and some node crash

flixhost

Member
Jan 8, 2020
19
0
6
27
Frankfurt am Main
flix-host.de
Hello guys,



We do have some customers that like to run proxmox in proxmox.

I don't exactly know why, but i kind of don't think this is actually good.

Does someone have some long-term experience with it?
Something that is important?

I've witnessed a node crash recently.

I think that is some relevant log of the cause or fault that happened here:

Modules linked in: veth act_police cls_basic sch_ingress sch_htb nfsv3 nfs_acl ip6t_REJECT nf_reject_ipv6 ipt_REJECT nf_reject_ipv4 xt_mark xt_set xt_physdev xt_addrtype xt_comment xt_multiport xt_conntrack nf_conntrack nf_defrag_ipv6 nf_defrag_ipv4 xt_tcpudp ip_set_hash_net rpcsec_gss_krb5 auth_rpcgss nfsv4 nfs lockd grace fscache sctp tcp_diag inet_diag bonding ebtable_filter ebtables ip_set ip6table_raw iptable_raw ip6table_filter ip6_tables iptable_filter bpfilter nf_tables softdog nfnetlink_log nfnetlink ipmi_ssif amd64_edac_mod edac_mce_amd kvm_amd kvm irqbypass crct10dif_pclmul crc32_pclmul ghash_clmulni_intel aesni_intel crypto_simd cryptd glue_helper pcspkr ast drm_vram_helper ttm drm_kms_helper drm fb_sys_fops syscopyarea sysfillrect sysimgblt joydev input_leds ccp k10temp ipmi_si ipmi_devintf ipmi_msghandler mac_hid sunrpc vhost_net vhost tap ib_iser rdma_cm iw_cm ib_cm ib_core iscsi_tcp libiscsi_tcp libiscsi scsi_transport_iscsi ip_tables x_tables autofs4 zfs(PO) Oct 4 22:00:18 pve31 kernel: [7883285.315751] zunicode(PO) zzstd(O) zlua(O) zavl(PO) icp(PO) zcommon(PO) znvpair(PO) spl(O) btrfs xor zstd_compress raid6_pq libcrc32c ses enclosure hid_generic usbmouse usbkbd usbhid hid igb i2c_algo_bit ixgbe mpt3sas ahci xfrm_algo raid_class libahci dca xhci_pci mdio scsi_transport_sas xhci_hcd i2c_piix4

ct 4 22:00:18 pve31 kernel: [7883285.315294] WARNING: CPU: 11 PID: 15453 at arch/x86/kvm/mmu.c:2250 nonpaging_update_pte+0x9/0x10 [kvm]

Regards and thank you for the answers!
 
Hi,
Does someone have some long-term experience with it?
Well, not in production but I and other devs often use it a lot for development and testing.

Something that is important?
Ensure that nested virtualization is available, see https://pve.proxmox.com/wiki/Nested_Virtualization (note that wit PVE 7.x nesting feature should default to one, so you may not need to enable it explicitly on the module anymore).

ct 4 22:00:18 pve31 kernel: [7883285.315294] WARNING: CPU: 11 PID: 15453 at arch/x86/kvm/mmu.c:2250 nonpaging_update_pte+0x9/0x10 [kvm]
Hmm, what CPU model do you use in that system?
 
Hi and thanks for the reply!
Well, not in production but I and other devs often use it a lot for development and testing.
Yep thats also what i thought.
Ensure that nested virtualization is available, see https://pve.proxmox.com/wiki/Nested_Virtualization (note that wit PVE 7.x nesting feature should default to one, so you may not need to enable it explicitly on the module anymore).
Allredy checked.
Hmm, what CPU model do you use in that system?

1633622703506.png
 
Hmm, that Epyc CPU model is about as old as the 5.4 kernel, and I have some faint memories about support for those improving in newer ones, especially at more advanced things like KVM nesting. Maybe you can try to see if one of the following options provides some remedy: