mirror of
https://github.com/librenms/librenms.git
synced 2024-10-07 16:52:45 +00:00
caf1cd3707
Added stopping and starting RRDCached during the change of the rrd files step. This will prevent data loss and corruption of the rrd files.
102 lines
3.1 KiB
Markdown
102 lines
3.1 KiB
Markdown
source: Extensions/Fast-Ping-Check.md
|
|
path: blob/master/doc/
|
|
|
|
# Fast up/down checking
|
|
|
|
Normally, LibreNMS sends an ICMP ping to the device before polling to
|
|
check if it is up or down. This check is tied to the poller frequency,
|
|
which is normally 5 minutes. This means it may take up to 5 minutes
|
|
to find out if a device is down.
|
|
|
|
Some users may want to know if devices stop responding to ping more
|
|
quickly than that. LibreNMS offers a ping.php script to run ping
|
|
checks as quickly as possible without increasing snmp load on your
|
|
devices by switching to 1 minute polling.
|
|
|
|
**WARNING**: If you do not have an alert rule that alerts on device
|
|
status, enabling this will be a waste of resources. You can find one
|
|
in the [Alert Rules
|
|
Collection](../Alerting/Rules.md#alert-rules-collection).
|
|
|
|
## Setting the ping check to 1 minute
|
|
|
|
1: If you are using [RRDCached](http://docs.librenms.org/Extensions/RRDCached/), stop the service.
|
|
|
|
- This will flush all pending writes so that the rrdstep.php script can change the steps.
|
|
|
|
2: Change the ping_rrd_step setting in config.php
|
|
|
|
```
|
|
$config['ping_rrd_step'] = 60;
|
|
```
|
|
|
|
3: Update the rrd files to change the step (step is hardcoded at file
|
|
creation in rrd files)
|
|
|
|
```
|
|
./scripts/rrdstep.php -h all
|
|
```
|
|
|
|
4: Add the following line to /etc/cron.d/librenms to allow 1 minute
|
|
ping checks
|
|
|
|
```
|
|
* * * * * librenms /opt/librenms/ping.php >> /dev/null 2>&1
|
|
```
|
|
|
|
5: If applicable: Start the [RRDCached](http://docs.librenms.org/Extensions/RRDCached/) service
|
|
|
|
|
|
**NOTE**: If you are using distributed pollers you can restrict a
|
|
poller to a group by appending `-g` to the cron entry. Alternatively,
|
|
you should only run ping.php on a single node.
|
|
|
|
## Sub minute ping check
|
|
|
|
Cron only has a resolution of one minute, so we have to use a trick to
|
|
allow sub minute checks. We add two entries, but add a delay before one.
|
|
|
|
Alerts are only run every minute, so you will have to modify them as
|
|
well. Remove the original alerts.php entry.
|
|
|
|
1: Set ping_rrd_step
|
|
|
|
```
|
|
$config['ping_rrd_step'] = 30;
|
|
```
|
|
|
|
2: Update the rrd files
|
|
|
|
```
|
|
./scripts/rrdstep.php -h all
|
|
```
|
|
|
|
3: Update cron (removing any other ping.php or alert.php entries)
|
|
|
|
```
|
|
* * * * * librenms /opt/librenms/ping.php >> /dev/null 2>&1
|
|
* * * * * librenms sleep 30 && /opt/librenms/ping.php >> /dev/null 2>&1
|
|
* * * * * librenms sleep 15 && /opt/librenms/alerts.php >> /dev/null 2>&1
|
|
* * * * * librenms sleep 45 && /opt/librenms/alerts.php >> /dev/null 2>&1
|
|
```
|
|
|
|
# Device dependencies
|
|
|
|
The ping.php script respects device dependencies, but the main poller
|
|
does not (for technical reasons). However, using this script does not
|
|
disable the icmp check in the poller and a child may be reported as
|
|
down before the parent.
|
|
|
|
# Settings
|
|
|
|
ping.php uses much the same settings as the poller fping with one
|
|
exception: retries is used instead of count.
|
|
ping.php does not measure loss and avg response time, only up/down, so
|
|
once a device responds it stops pinging it.
|
|
|
|
```
|
|
$config['fping_options']['retries'] = 2;
|
|
$config['fping_options']['timeout'] = 500;
|
|
$config['fping_options']['interval'] = 500;
|
|
```
|