Availability checks and polling
The one-minute availability checks, the full reads on the polling schedule, and every field on Settings › Monitoring › Schedule, with defaults.
ConceptReviewed against the application source on October 12, 2026
Two kinds of reading
Full reads collect a device’s health and power, ports and endpoints, neighbors and parts on the polling schedule. Availability checks take one quick look at each device about once a minute: one SNMP uptime request or, for a switch read over SSH only, a connection to port 22 that ends before signing in, plus a ping where the appliance may send one. They exist so that a device that stops answering is known within about two minutes, however long a full cycle takes across thousands of devices.
Turn on the one-minute checks
- As an Administrator EX, open Settings › System › Features.
- Turn on Check every device every minute. The checks then run and record what they see, without opening outages.
- When you trust what they record, turn on Open “not answering” outages from the one-minute checks. A device that misses two quick looks about 15 seconds apart becomes an outage within about two minutes, and the outage closes at its first answer. This also turns the checks on.
- Optionally, turn on One outage for a switch and the switches behind it; see Work an outage.
All three are off until you turn them on. Full reads keep deciding everything else: stack members, power supplies, uplinks and restarts. Priority sites still wait the minutes their site type sets before an outage opens.
Settings › Monitoring › Schedule
The Schedule tab has three cards. Automatic polling holds Enable automatic polling and Polling interval (minutes), 5 to 1440, for every site that is not a priority site. Availability checks and Full reads hold the fields below. Changing them needs an Administrator EX.
| Field | Default | What it does |
|---|---|---|
| Availability check interval | 60 seconds (30 to 300) | How often each device is asked for one value to see that it answers. A device that misses is asked again about 15 seconds later before it counts as not answering Cenovel. |
| Availability checks at once | 128 (16 to 512) | How many devices are checked at the same time. Each check is one small request, so this can be high; it bounds how long a round takes when many devices do not answer. |
| Timeout | 4,000 ms | How long to wait for an answer. SSH sign-in waits at least 10 seconds. |
| SNMP full reads at once | 16 (1 to 64) | How many switches a full SNMP read (ports, parts, power and neighbors) runs on at the same time. |
| SSH sessions at once | 8 (1 to 32) | How many switches are read over SSH at the same time, one session each. Each switch allows only a few sessions. |
| Table reads per switch | 4 (1 to 8) | How many tables one full read asks a switch for at the same time. A switch that answers slowly is read one table at a time. |
An appliance upgraded to Opal with no value saved reads 16 SNMP switches and 8 SSH switches at once. A value saved for the retired Switches at once setting carries over to both.
Priority sites
In Settings › Inventory data › Locations › Site types, a type whose outage urgency is Priority has its own Read priority sites every (2 to 60 minutes) and Open an outage after. Its outages are critical, and their alerts start with URGENT.
Tune with care
- Raise the figures only while devices and the network keep up: a higher number finishes a cycle sooner and puts more on the network at once.
- If many devices stop answering together, Cenovel may have lost its own path to the management network. It then says “Cenovel can’t reach the management network”, holds device outages, and calls nothing down until it can read again.
- Devices behind a site collector are checked by the console for now; they must be reachable from it.
See also: Device reads and monitoring · Outages and planned maintenance
Describes the release being prepared for deployment. Check the behavior on your installed release.