Age | Commit message (Collapse) | Author |
|
* makefiles install configs in /usr/lib/netdata/conf.d; #4182
* stock health config in /usr/lib/netdata/conf.d/health.d
* unit test path concatenation
* simplified health file management
* use stream.conf from stock config if it does not exist in /etc/netdata
* indicate loading of user config in function call
* load netdata.conf from stock dir if not found in /etc/netdata
* added NETDATA_USER_CONFIG_DIR
* provide defaults before loading config
* charts.d uses stock files
* fping now uses the stock config files
* tc-qos-helper.sh now uses stock configs
* cgroup-name.sh now uses stock configs too
* simplified cgroup-name.sh for user and stock config
* alarm-notify.sh uses stock configs too
* simplified fping plugin configs loading
* simplified tc-qos-helper.sh configs loading
* added error handling to charts.d.plugin
* apps.plugin used stock configs
* generalized recursive double-directory configs loading
* statsd uses stock configs
* node.d.plugin uses stock configs
* compile-time decision of netdata default paths for all files
* makeself cleans up old stock config files from user configuration directories
* fixed makeself typo
* netdata-installer.sh removes stock files from user configuration directories
* python.d.plugin user/stock configs update
* cleanup stock config files from /etc/netdata, only once
* python.d.plugin log loaded files
* fix permissions of stock config files and provide an "orig" link for quick access
* create help link on stock configs migration for static installations
* create user config directories
* example statsd synthetic charts now state they are examples
* updated configs.signatures
* spec file
* fixes in spec file
* fix typo
* install netdata after cleaning up stock configs from /etc/netdata
* python.d: add cpuidle stock conf
|
|
|
|
|
|
* replaced referenced to firehol github org with netdata github org
* increased versions of js files
* added new docker hub badge netdata/netdata and restored firehol/netdata
|
|
|
|
* pyhton.d.plugin: respect update_every in jobs
* pyhton.d.plugin: run gc.collect every 300 secs in main thread
* pyhton.d.plugin: gc collect work around comment
* python.d.plugin: do not run all modules debug
* python.d.plugin: make gc.collect runs optional, add options to python.d.conf
|
|
|
|
|
|
|
|
The Simple Notification Service (SNS) is a reasonably simple message
broker service provided by Amazon as part of it's AWS offerings.
SNS utilizes the concept of 'topics' (similar to Netdata's concept of
'roles' for notifications) to control message routing. Any given topic
may have any number of subscribers of any of the following types:
* Email addresses.
* Phone numbers for SMS.
* HTTP or HTTPS web hooks.
* AWS Lambda endpoints.
* AWS SQS endpoints.
* Mobile applications (via various native push notification services).
Topics are rpersented by Amazon Resource Names (ARN), which are a special
type of URI.
Unfortunately, properly signing requests to AWS endpoints is a serious
pain in the arse, so we pretty much have to use the CLI interface.
Because of the inflexibility of this tool, setup is somewhat painful.
THis uses topic ARN's directly as recipients. SNS does not support
delivery to multiple topics in bulk, so a separate call is required for
each topic ARN. Users are provided with the ability to customize the
message format used for the notifications.
|
|
|
|
|
|
https://github.com/tioumen/netdata into ms_team_notification_support
Merging with @ktsaou one
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Add fleep.io notification support.
|
|
Add rocket.chat notifications
|
|
Licensing cleanup.
|
|
This adds support for sending alarm notifications to fleep.io
conversations via webhooks. It uses the JSON interface to POST a simple
message containing information about the alarm, and by default uses the
host that raised the alarm as the sender name.
Recipients are specified via the last component of the webhook URL (the
only part that's different for each webhook).
|
|
|
|
|
|
|
|
This tags all the files that I can clearly identify as being part of
Netdata and not a local copy of a third party package that we're
redistributing.
|
|
This adds a new configuration option in health_alarm_nofity.conf to
insert In-Reply-To and References haders in email notifications. The
contents of these headers are composed in a deterministic way from the
chart name, alarm name, and host name.
This allows email clients to sort netdata alert emails into groups based
on the combination of chart, alarm, and host, which can help
significantly when dealing with a large number of alerts. It also
provides something for scripts that might process and filter thes emails
to match on that's a bit easier to parse than the subject line.
We default to not enabling this so that people who have their e-mail
clients configured for threaded reading don't see an unexpected change
in behavior.
Strictly speaking, this is technically a misuse of the headers, as they
are supposed to only reference the Message-ID header of an existing
e-mail. However, I know of no e-mail clients that enforce standards
compliant behavior (as it would break threading picked up partway
through a thread), so we should be safe.
Minimally tested (using the test-alert.sh script and running for a short
while on a server where I synthetically induced lots of alerts).
|
|
4.4.18 and curl to 7.60.0
|
|
|
|
|
|
sleep if builtin sleep fails
|
|
|
|
|
|
invalid use of shell -eq with strings in alarm_notify.sh syslog support
|
|
|
|
The original commit unintentionally capitalized the name of the
`date_format` variable in plugins.d/alarm-notify.sh, which caused the
new date handling support to not work at all (but thankfully did not
break any existing functionality).
This corrects that issue, so that the new date handling support works
correctly.
|
|
Add syslog support to alarm-notify.sh.
|
|
This adds support to alarm-notify.sh to report dates in alert
notifications using an arbitrary format string.
The current default behavior (just using the default format fromm
`date`, which is simply the ANSI C asctime() default) is preserved.
In the event that the user specified format string doesn't work, the
script will fall back to using the default output of the date command.
THe config file includes comments listing a handful of widely used
internet date formats as examples.
This also removes the explicit quoting around the shell expansion used
to get the date, as it's unneeded (shell expansion results done with $()
is implicitly treated as a single string), and makes it very hard to
get this working properly.
|
|
THis adds the chart name to the syslog messages produced by
alarm-notify.sh.
After this, log messages will now look like:
prefix WARNING on wild-karde at Tue Apr 3 09:00:00 EDT 2018: disk_space._ out of disk space time = 5h
|
|
- Changed to produce "0" exit code on success
- Exit code "1" is produced only if there is any error spotted
|
|
|
|
This adds support to syslog notifications to have per-role
configuration. This includes specifying facilty and log-level per-role,
target servers per-role, and a message prefix per-role, possibly
multiple times. This is done by specifying each combination of
configuration as a 'recipient' in the role recipients variables using
the following format:
[[facility.level][@host[:port]]/]prefix
As part of this, the default log level has been changed to depend solely
on the type of alert, crit for CRITICAL, warning for WARNING, and info
for everything else (most notably CLEAR), and the end-user configuration
and the user configuration of these default levels has been removed.
In addition, the message format has been changed to look like the
following:
Apr 3 09:00:30 localhost netdata: prefix WARNING on wild-karde at Tue Apr 3 09:00:00 EDT 2018: out of disk space time = 5h
|
|
|
|
This uses the logger command (found on all modern Linux and BSD systems)
to allow alarm-notify.sh to send netdata alarm messages to a local
or remote syslog server. By default, logging is done to the local6
facility at a warning log level, with critical alarms being logged at
the err level.
This provides a mechanism of easily generating a running log of all
netdata alarm events on most standard UNIX systems, and when used with
full-featured logging daemons like rsyslog or syslong-ng can be used to
archive such messages in a wide varity of alternative formats.
There are a few limitations currently:
* It can only log to one location. This can be easily worked around in
most syslog daemon's configurations, as almost all of them support some
form of output multiplexing.
* The default settings will usually not get the results most people
would want, but because of the sometimes very significant differences
in default syslog configurations from distribution to distribution,
there's not really any way to make the defaults any saner.
* While it is possible to log directly to a remote syslog server, there
is no quick and easy wet up for this, it requires the user to manually
compose the extra options that need to be passed to the logger command.
Syslog messages generated by this will look something like this:
Apr 3 09:00:30 localhost netdata: Netdata notification on wild-karde at Tue Apr 3 09:00:00 EDT 2018: WARNING, out of disk space time = 5h
|