Changed
- Ensure all telegraf metrics are timestamped in microseconds. Because output to
telegraf is buffered, we can't rely on telegraf doing the timestamping
(multiple samples get timestamped with the same or very similar times), so we
add a timestamp to every metric line as soon as we see it. ntpmon_info
metric in telegraf mode split into
ntpmon_resident_set_size_bytes
,ntpmon_virtual_memory_size_bytes
, and
ntpmon_uptime_seconds
in prometheus mode, with tags only on
ntpmon_uptime_seconds
, to tidy metric names and reduce tag cardinality when
used with prometheus. These retain their existing names in telegraf mode.- Make messages from log file tailer tidier and more consistent.
- Correctly flag
/etc/default/ntpmon
as a configuration file in the Debian
package.
Full Changelog: v3.0.7...v3.0.8