- Connect the hyperlink signal only when configuring the plugin.
Once we'll be able to configure the plugin from the GUI, we'll also
need to disconnect it first.
- Allow to configure server (UI) keepalive and keepalive timeout values.
- Increase spinBox values using the single-step defined value.
- Translate left panel section names.
Main window split into two areas:
- a left panel listing the main configuration sections.
- a right panel containing the configurable options.
We were running out of space, so this should help to organize all the
options a bit better.
When changing the firewall policies, if there're more than node
connected and node selected is All, ask the user to confirm the changes.
To avoid surprises.
- Use text instead of icons, because in some themes the - icon is not
represented as "-" character.
- Put the + button on the right, and - on the left.
Closes: #1473
This option was causing a reaaally annoying behaviour:
When you scrolled horizontally the view and clicked on a row,
the view would automatically scroll to the left.
When clicking on a node, the GUI starts a background task to obtain the
system stats of the daemon (ram, swap, uptime, load avg, ...).
We were not handling the monitoring/unmonitoring actions properly.
When a node connected to the gui for the first time, we were not
updating the db accordingly. We were doing it when populating the
stats or adding events.
Unfortunately, not all fields are available when a node subscribes to
the GUI, they're sent when the node posts new events. Because of this,
some fields may appear blank until the node send new events.
It'll be fixed in future releases.
When showing/hiding the main window by clicking on the systray icon, if
the user was reviewing the connections of a node before hiding the
window, the node table was not restored again.
We changed a few weeks ago how the window is shown (showEvent() ->
show()), so this is an unexpected error from that change.
- allow to configure max number of nodes (incoming connections).
If the server address is a unix socket, it'd make sense to limit it
to 1.
- allow to configure server workers.
By default ThreadPoolExecutor max_workers value is calculated based
on the number of CPUs: min(32, os.cpu_count() + 4).
When the GUI (server) is configured to manage multiple nodes, if
there're not enough server workers (for example because there's
only 1 CPU, thus ~6 threads), it can end up freezing, blocking new
requests (new nodes, new stats, new rule requests, etc).
The server uses one worker per node to handle notifications, and
it's not released until the node exits.
Every second, the nodes send the statistics to the server, which also
consumes one worker for at least one second.
Additionally the server may receive new rule requests and alerts,
which also consume one worker each temporarily.
So if there're 6 nodes, the server may need ~3 workers per node.
Related: #1453