Skip to content

Collimo 0.3.0

Choose a tag to compare

@sf-grumpygeek sf-grumpygeek released this 05 Nov 18:16
· 2 commits to main since this release

Before you upgrade

One change needs a look first.

  1. Allocation forecasts no longer raise alerts on their own. The runway for CPU, memory and disk allocated to VMs is now shown for planning only, and what is actually in use is what alerts. If you relied on an allocation alert, for example because you never overcommit memory, set CAPACITY_ALLOC_ALERTS=alloc_mem (any of alloc_cpu, alloc_mem, alloc_disk, comma-separated). Forecasts

Four new variables, all optional and empty by default: CERT_WATCH, DOMAINS_WATCH, CAPACITY_ALLOC_ALERTS and CAPACITY_WAN. Settings → Stack environment writes them for you. Variables

New

  • A Renewals tab. Every certificate Collimo already sees (https sources, https up-checks and consoles), plus any you add in CERT_WATCH, your domains over RDAP if you list them in DOMAINS_WATCH, and vCenter licenses, soonest first under a year planner. A certificate warns once a third of its own lifetime is left. Collimo reads the dates and never renews anything. Free in every edition. Renewals
  • Resilience: what goes down with it. What fails with each switch, UPS, PDU and host, biggest first, and the single points of failure, such as a host whose every power supply is on one UPS. It's built from what Collimo already reads: uplinks and LLDP, the switch port each host plugs into, and NetBox's power cables. A Power table lists what feeds every switch, PDU, host and server, and lets you set a feed by hand where NetBox doesn't say. Anything Collimo can't see is listed as not known, never guessed. Resilience
  • A topology map of the same graph, drawn as rack units, on the Resilience tab. Pick a unit to see what fails with it. Topology map
  • Cluster consistency, on the Virtualization tab: hosts in one cluster whose build, BIOS, BMC firmware, MTU or VLANs differ from the rest. Cluster consistency
  • Placement advice, on the Virtualization tab: which VMs to move to even out a Proxmox VE cluster, or a vCenter cluster with DRS off, and why any are held. It's advice only; Collimo never moves a VM. Placement advice
  • UPS battery aging: each battery's health, trend and age from its runtime at its load, with a warning before runtime falls under 10 minutes. UPS battery aging
  • Network capacity, on the Capacity tab: uplinks and WANs at their 95th percentile, and PoE budgets. Set CAPACITY_WAN to judge a WAN against its contracted speed instead of its best speedtest. Network capacity
  • Rack cooling, on the Capacity tab: each rack's inlet air from its servers' intake sensors, hot racks, and the daily trend. Rack cooling
  • Audit evidence packs, from Settings: a zip for an auditor, signed by your install's own key, which collimo -verify-pack checks. Evidence packs
  • Intel AMT, a new source kind for machines without a BMC, so cluster consistency can compare their BIOS. Intel AMT
  • Every section folds to its heading, on every tab, and a folded section opens again when a new alert shows up in it. Expand all and Collapse all are kept through a reload. First look

Changed

  • Right-sizing says when a VM is too small as well as too big, and vCenter VMs are no longer sized on active memory. Reclaim
  • VLANs are read for UniFi ports that follow a port profile, and for Custom ports.
  • Hosts are placed on switch ports by LLDP neighbors the UniFi client list doesn't name.
  • Fixes to the Balance panel, the planning colors, Cooling, notification groups and the diagnostics bundle.

Upgrading

Set COLLIMO_VERSION=0.3.0 and redeploy. Read Updating Collimo first if you haven't kept a copy of SETTINGS_KEY and BACKUP_KEY.