mrcharles
1f44a193a2
- cnodes 102/705/603 moved to TS5-SSD storage by founder (etcd apply 4-5s -> <1s) - wnode-tsys5 slot retired; ultix-streaming joins in its place (5 workers) - ultix-offstage unreachable; stays excluded until back on tailnet [#367][#368] https://projects.knownelement.com/issues/368
63 lines
3.4 KiB
Bash
63 lines
3.4 KiB
Bash
#!/usr/bin/env bash
|
|
#
|
|
# k8s/proxmox-tuning.sh — apply ultix-style Proxmox tuning to k8s VMs
|
|
#
|
|
# Codifies the config applied 2026-09-01 during the #367 rebuild:
|
|
# - disk: ssd=1,discard=on (guest sees non-rotational; qcow2 trims)
|
|
# - net: queues=2 on both NICs (multiqueue; guest activates via ethtool)
|
|
# - cpu: cpuunits so etcd/control plane wins host CPU contention
|
|
# - boot: onboot=1 + startup order (cnodes before wnodes)
|
|
# - balloon was already 0 on every k8s VM (etcd memory predictability)
|
|
#
|
|
# Pending items (ssd/discard, queues) activate at the VM's next restart.
|
|
# Reboot waves are SERIAL with a health gate between hops — never batch.
|
|
#
|
|
# DRY RUN by default. Apply with RUN=1. Run on each Proxmox host, or via
|
|
# PROX_HOST=<host> bash tests/remote.sh prox 'bash -s' < k8s/proxmox-tuning.sh
|
|
#
|
|
set -uo pipefail
|
|
RUN="${RUN:-0}"
|
|
|
|
# vmid|cpuunits|startup|scsi0-line|net0-line|net1-line
|
|
# scsi0/net lines are the FULL desired config (Proxmox replaces wholesale).
|
|
TUNINGS=(
|
|
"102|4000|order=20,up=180|TS5-SSD:102/vm-102-disk-0.qcow2,cache=writeback,iothread=1,ssd=1,discard=on,size=32G|virtio=BC:24:11:CB:97:10,bridge=vmbr0,firewall=1,queues=2|virtio=BC:24:11:47:13:08,bridge=datanet,queues=2"
|
|
"705|4000|order=20,up=180|TS5-SSD:705/vm-705-disk-0.qcow2,cache=writeback,iothread=1,ssd=1,discard=on,size=32G|virtio=BC:24:11:40:25:F8,bridge=vmbr0,firewall=1,queues=2|virtio=BC:24:11:AF:3D:89,bridge=datanet,queues=2"
|
|
"603|4000|order=20,up=180|TS5-SSD:603/vm-603-disk-0.qcow2,cache=writeback,iothread=1,ssd=1,discard=on,size=32G|virtio=BC:24:11:38:C0:58,bridge=vmbr0,firewall=1,queues=2|virtio=BC:24:11:1F:F1:E1,bridge=storagenet,queues=2"
|
|
"313|2000|order=40,up=120|local-lvm:vm-313-disk-0,iothread=1,ssd=1,discard=on,size=300G|virtio=BC:24:11:EE:7E:7B,bridge=vmbr0,firewall=1,queues=2|virtio=BC:24:11:DF:C2:E9,bridge=datanet,queues=2"
|
|
"601|2000|order=40,up=120|local-lvm:vm-601-disk-0,iothread=1,ssd=1,discard=on,size=32G|virtio=BC:24:11:FA:6E:B5,bridge=vmbr0,firewall=1,queues=2|virtio=BC:24:11:39:7D:0B,bridge=storagenet,queues=2"
|
|
"701|2000|order=40,up=120|S2:701/vm-701-disk-0.qcow2,cache=writeback,iothread=1,ssd=1,discard=on,size=32G|virtio=BC:24:11:30:B8:07,bridge=vmbr0,firewall=1,queues=2|virtio=BC:24:11:C8:0E:8A,bridge=datanet,queues=2"
|
|
"905|2000|order=40,up=120|S2:905/vm-905-disk-0.qcow2,cache=writeback,iothread=1,ssd=1,discard=on,size=32G|virtio=BC:24:11:EE:2B:B6,bridge=vmbr0,firewall=1,queues=2|virtio=BC:24:11:9B:57:06,bridge=datanet,queues=2"
|
|
)
|
|
|
|
apply() {
|
|
local vmid="$1" cpuunits="$2" startup="$3" scsi0="$4" net0="$5" net1="$6"
|
|
if [ "$RUN" = "1" ]; then
|
|
qm set "$vmid" --cpuunits "$cpuunits" --startup "$startup" \
|
|
--scsi0 "$scsi0" --net0 "$net0" --net1 "$net1"
|
|
else
|
|
echo "DRY: qm set $vmid --cpuunits $cpuunits --startup $startup --scsi0 $scsi0 --net0 $net0 --net1 $net1"
|
|
fi
|
|
}
|
|
|
|
for t in "${TUNINGS[@]}"; do
|
|
IFS='|' read -r vmid cpuunits startup scsi0 net0 net1 <<< "$t"
|
|
# Skip VMs that live on other hosts (qm set errors on unknown VMID)
|
|
if ! qm status "$vmid" >/dev/null 2>&1; then
|
|
echo "skip: VM $vmid not on this host"
|
|
continue
|
|
fi
|
|
echo "== VM $vmid =="
|
|
apply "$vmid" "$cpuunits" "$startup" "$scsi0" "$net0" "$net1"
|
|
done
|
|
|
|
echo
|
|
if [ "$RUN" = "1" ]; then
|
|
for t in "${TUNINGS[@]}"; do
|
|
vmid="${t%%|*}"
|
|
qm status "$vmid" >/dev/null 2>&1 || continue
|
|
echo "--- pending VM $vmid ---"
|
|
qm pending "$vmid" | grep -E "^(new|cur) (scsi0|net0|net1|cpuunits|startup)" || true
|
|
done
|
|
fi
|