Stop treating the same disk alert as a new incident: reconstruct the growth trend, project
when the volume fills, and produce an options brief — clean up, expand, re-architect.
Needs the RMM for current state.
1. Current state: the device details for capacity and free space per volume; documentation
(IT Glue / Hudu) for what the volume holds (file shares, databases, backup targets) —
the workload drives growth character and expansion options. Resolve org -> device
without stopping to ask; a class filter is not evidence.
2. Reconstruct the trend — the RMM exposes point-in-time readings only. Harvest dated
points from the device's alerts and activity (each threshold alert is a dated reading)
and from earlier disk tickets recording free-space figures. Three or more make a usable
trend; fewer and the honest output is "insufficient history — start recording,
re-forecast in N weeks", plus the current-state brief.
3. Compute growth honestly: rate per month from the points you have, a projected-full date
at that rate, a sensitivity line ("at 1.5x the rate, full by <earlier date>"). Deduct
known one-off events (a migration, a cleanup) rather than letting them distort the
trend — say when you did. State the forecast as a range, never a date certain.
4. Separate reclaimable from organic growth: old user profiles, stale snapshots, log bloat
and duplicate backup copies are reclaimable (quick wins go to disk-space-remediation);
steady LOB data growth is organic, and cleanup only rents time against it. Quantify the
runway cleanup buys against the growth rate.
5. Expansion options, matched to the platform from documentation: extend the volume or
datastore (if the underlying storage has headroom), add disks or a shelf (check slots
and warranty/EOL — expanding a dying unit is money misspent; see
hardware-refresh-forecast), tier or archive cold data, or move the workload (larger
NAS, SAN, cloud). For each: effort class, disruption (window needed?), and runway
bought at the observed rate.
6. Output current state, the trend with its data points, the projected-full range, the
reclaimable-vs-organic split and the options, as a note (apply the PSA Note Discipline
base skill — plain text, no markdown or emojis). Open a ticket for the account team
when the projected-full date is inside the client's procurement lead time.
Guardrails: a forecast from fewer than three dated readings is labelled low-confidence or
not issued — two points are not a trend. Cleanup recommendations name what the files are
believed to be and require verification before deletion; this skill deletes nothing. No
hardware model or price quotes from memory — sizing and pricing is the account team's.
Never forecast a backup-target volume as organic growth — retention settings drive it, not
users; say when growth is policy-driven.
Magic Library