NOTE: The existing alerts configuration of the users will be discarded.
Some code has been added to make the current gui on/off toggle work.
It is marked with the following comment:
-- TODO remove after implementing the new gui
This was caused by probe.getTimeseriesMenu being called at every page load by the timeseries
code to load the schemas. probe.loadSchemas is now called instead.
- Add title label in hash table utilization gauges
- List Hash Tables charts and Periodic Activities charts in a single page
- Replace Periodic Activities max duration column with time utilization percentage
- Add api to add a straight line in the charts
- Show max duration time as a straight line in Periodic Activities charts
- Move dropped flow calls charts to the Peridic Activities page
- Fix expert view number of calls 0
- Fix interface/host/network trigger alert label
- Move chart icon to separate column in redis and influxdb pages
- Fix interface menu selection in Hash Tables and Periodic Activities
- Highlight issues with a warning triangle in Hash Tables and Periodic Activities
- Fix some page navigation issues with extra_params
- Fix wrong schema graph options passed to nv_graph_utils
Using alerts with default values caused the already_triggered function to miss the existing alert, which
was automatically released. Moreover, the alerts_api.releaseEntityAlerts passed a nil cur_alerts,
breaking releaseEntityAlerts.
This provides a ~10x speedup by performing the JSON serialization work in C
(and thus avoiding Lua->C overhead). This also implements two in-memory alerts
queues (one for sqlite and one for the notifications) in order to reduce Redis load.
Alerts queue are now global instead of per-interface as there is only 1 dequeing thread.
