nothing open - no alert is active, pending or clearing
Device health
Locations
location
up
down
pending
Applications
application
up
down
pending
Service health
checks
ok
warning
failing
idle
Sites
site
location
MOS
vs prev
24 h low
loss / jitter
voice ran
Mbps to / from
vs prev
loaded ms
bandwidth ran
Web and TCP
check
device
result
ok 24 h
cert
last run
Top 10
Received
device
interface
total
peak
vs before
Transmitted
device
interface
total
peak
vs before
Errors
device
interface
errors
vs before
Discards
device
interface
discards
vs before
CPU
device
sensor
average
peak
vs before
Memory
device
sensor
average
peak
vs before
Totals are each hour's average rate times the hour, from the
interfaces' own counters - a missed poll loses no traffic. Hours a device did not
answer at all, or rebooted through, count as nothing; a row that saw less than 95% of
the window says how much it saw. "vs before" compares with the window just before this
one. Memory is used over size as the device reports them, and some devices count
caches as used.
Interface traffic report
chosen:none yet - pick a device above and tick its interfaces
Availability is the share of the once-a-minute pings answered; pings
inside a maintenance window are left out. Measured is how much of
the period anyone was looking. Outages come from the ten-second
sweep: down a minute or more is an outage, less is a blip, and one
that began in maintenance is counted apart.
Search
export
state
rows
size
started
time
severity
host
app
message
Alerts
severity
state
alert
value
peak
threshold
raised
last seen
delivery
all quiet - and the watchdog is what makes that different from not looking
Recently cleared
severity
alert
peak
duration
cleared
reason
Reachability
when
device
transition
rtt ms
no transitions recorded - a fleet that holds its state writes nothing here
Delivery log
when
event
channel
result
detail
Health
thread
worst, 15 min
worst since start
p99
over 50ms
lane
in flight
capacity
queued
Add device
One line per service: an IP address, then a space, then the name you want on
the wall. These are watched by ICMP alone - up, down and latency, with alerts
and maintenance windows like any other device.
What answered
address
name
interfaces
tracked
sensors
location reported
result
Locations these devices report
use
reported by the device
devices
save as
Untracked ports stay listed on the device page and can be
ticked later. Manual mutes link down only - errors, discards and utilization
still alert - and each port is switched on from its row (uplinks, host ports).
A device added without an answer tracks by the usual rule when it first answers.
Devices
device
a guest PC, parlor station, HTPC: off is a state, not a
fault. While set, device-down never raises - the device renders "off"
and keeps being polled. Saves when toggled.
nothing on this device raises - down, interfaces, sensors.
It is still polled and charted, and anything open clears within a few scans.
Syslog and trap rules are not affected. Saves when toggled.
link down alerts only on the ports you turn on with their
"alert on" button - for uplinks and host ports on a switch. Errors, discards and
utilization are unchanged. Saves when toggled.
a warn when the device or its SNMP agent restarts, open
for 30 minutes and cleared without an email. Never for a restart inside a
maintenance window, or for a transient device. Saves when toggled.
a warn when its OS, hardware model or boot firmware changes, open
for 30 minutes and cleared without an email. Never for a change inside a maintenance
window, or for a transient device. Every change is on its timeline either way.
Saves when toggled.
a warn when a temperature that follows the room runs well above
its own last week, or several in its location do at once - a failed fan or a failed AC
that every fixed threshold misses. Each card says whether it takes part. Saves when toggled.
notes
severity
state
alert
value
threshold
raised
Events
when
what
detail
Services
seconds each way, marked
Mbps each way - proves the site can still get that much
:00 and:00, this box's time - it fills the link while it runs
seconds (at least 30), time out aftersecondsseconds
Errors and discards
interface
description
speed
state
in
out
errors/s
discards/s
last reading
Maintenance windows
A window withholds notification for a named set until it
ends. The alerts still raise, stay on every page and the wall marked
"in maintenance", and anything still active when the window closes is
delivered on the next pass. The watchdog is never suppressed. A
repeating window - a nightly backup, a daily reboot - opens at its local
time on its days and closes after its minutes, every time.
covers
when
until
note
by
none - notifications flow normally
Notify policies
A policy is standing: alerts for the named set still raise
and stay visible everywhere, marked "unpaged - policy", and are never
delivered while the policy exists. Use it for Dev/Test and Non-Prod
groups you watch but never want paged about. Dropping a policy delivers
anything still active on the next pass. The watchdog is never covered.
covers
note
by
since
none - every alert pages normally
My account
At least 8 characters. This browser stays signed in; any other
session of yours is signed out, so a password someone else learned stops working
everywhere at once.
Retention
Growth
What happens next
What it did last
job
last outcome
last run
last success
runs
failures
Partitions
partition
table
rows (est)
size
free-text index
Settings
Environment only - there is no config store, so these
change by editing the unit and restarting.
variable
value
what it does
Thresholds
The shipped defaults are a starting point. An override replaces them
for one sensor (by its code), for one kind on one device,
or for that kind everywhere. A muted override suspends the rule
without touching polling or history. Narrower always wins: sensor
beats device beats kind beats default. Set overrides from a sensor's
card on its device page; this is the fleet-wide view.
scope
kind
target
warn
crit
state
note
Defaults
Syslog and trap alert rules
Raise an alert when an incoming syslog message or SNMP trap
matches a pattern - "authentication failure" in a log line, or a linkDown
trap's OID. The alert raises the moment a matching message arrives, names the
device that sent it, and clears on its own once no matching message has arrived
for the "clears after" time. Rules run on every message received, so keep
patterns specific.
rule
pattern
source
severity
clears after
enabled
s
Group alerts
One alert for a location or application when enough of its
devices are down at once, instead of an email for each. Tick a group to watch it.
It trips when at least the percent shown of its devices are down - counting only
devices whose status is known - and at least the minimum number, so a small group
cannot trip on one failure. While the group alert is open, its devices' own alerts
still raise and show, but their emails are held: the group's email names them.
Any still down when the group recovers are emailed then. Transient and muted
devices are not counted.
Outside services works the same way over the service checks that
count as outside - those reaching a public address, or marked so. When enough
of them fail at once, the likeliest cause is RSCanvas's own internet, and one
alert says so instead of one for each check.
watch
group
up
down
trips at
and at least
Credentials
A profile is a named SNMP credential the collector sends to every
device that names it. Use read-only communities and users:
if this store is ever compromised, the attacker should be able to
read your devices, not change them. Secrets are encrypted at rest
and never shown again once saved.
profile
version
secret
devices
updated
Create a profile
Users
username
role
created
last sign-in
Add a user
Boards and displays
board
collection
group
drift
addresses
displays
auto fits the display's own screen (the field set fixes the
tile height, so the arithmetic is exact); a number forces it;
both blank = drawn board from coordinates
the board's own groups keep the order they were declared in;
the other axis sorts its sections by name
Generate a board from a group
Displays for
label
created
last used
state
no displays yet - mint a token below and
open its URL on the screen that should show this board
