1.6.242.0901 Release Notes

Each switch port now shows what is plugged into it. Monitoring targets can be grouped by service. Ping is now sent from the PC running Manager, and 7.6 seconds of agent startup time is gone.

What is plugged into each switch port

Server monitor · Target detail

Each port on a registered switch now shows the address and manufacturer of the device attached to it. Devices sitting beyond an upstream switch are marked separately, since they are not plugged into that port directly. The table a switch keeps of attached devices is cleared after five minutes, so we read it before it disappears. The manufacturer is shown only for ports with a single device, because with several there is no way to tell whose it is.

Monitoring targets grouped by service

Server monitor · Target list

Group the targets that belong to one service, such as a shop's web, database and domain, and watch them together. Create a group from Add group in the list toolbar and drag targets into it. Use even one group and grouping becomes the outer axis; use none and the list stays sorted by type as before. An empty group still appears in the list.

Switch port figures that did not match reality

Server monitor

The port list and grid now count physical ports only; turn on Include virtual to see virtual interfaces. Korean descriptions written on the port show correctly, and ports with no link leave the speed blank. A 24-port switch with two cables reading 4/26 UP was counting the switch's own management interface.

Ping now sent from the Manager PC, not the server

Node list · right-click action

The agent on the PC running Manager sends the ping. The previous package described the server sending it, but on sites spread across the country a signal from the central server to an endpoint is often blocked by a firewall, so that approach did not hold. The feature is not gone; only the sender changed.

Faster agent start

Agent

Even with nothing to delete, clearing the previous boot's data took 6.4 seconds and a supplementary file lookup another 1.2. Both queries were corrected and that time is gone. Storage housekeeping that used to run right at startup was also deferred.

Symptoms raised wrongly or missed

Symptoms · Agent

Fewer entries come from ordinary browser traffic, and one-directional failures such as a proxy outage are now caught. Disk delay, boot delay and clock change symptoms fire exactly as the rules describe. A size comparison was being evaluated as text rather than numbers, so 900 counted as larger than 1000; that is fixed.

External and internal IP as two columns

Node list

Turn on External IP and Internal IP separately in the node list column picker. Each column shows only its own value and stays blank when there is none. Column settings you already saved expand into the two columns on open, keeping their shown and hidden state.

Same PC registered as a new node after reinstall

Agent

The order in which network adapter addresses are used to identify a PC is now fixed. This prevents a PC from appearing as a different one when Windows happens to reorder the same adapters.

Symptom history that survives a reboot

Agent

The screen shows only the current boot's symptoms, but the history of previous boots is now kept for 30 days. The agent used to delete them at startup. Some judgements, such as whether the motherboard battery is dead, need history spanning several boots.

PCs with recovery software failing to reach the server

Agent

In a previous version a damaged settings file left the agent unable to read its own identifier. The hand-written file handling was removed in favour of storage that Windows manages, which Windows repairs even if power is lost mid-write.

Remote control reachable without permission

Settings › Permissions

There was no place where the permission was checked, so a role with remote control switched off could still connect. Now the connection does not start, the reason is shown, and no reconnection is attempted in the background. The Manager session is sent along so the server can tell who requested it, and that value also fills the requester in the access log.

Process I/O reading up to 180 times off

Load · Processes

The same process could read up to 180 times differently between the chart and the list, inflated by the measurement interval. Chart scales that stretched regardless of the data now follow the real maximum, and the daily spread bar, which held a running maximum and never came back down, now shows that day's value.

Switching axis straight from the symptom detail

Symptom detail

In Symptom information, press the type, source or cause chip and the list regroups by that axis. The chip is inert for the axis you are already on or when the value is empty, because moving to an empty value would leave the list blank.

AI diagnosis

Team Lead Jung

The confirmation dialog now appears only for commands that change a setting or force a process to end. Previously a single web search in the conversation made every following read-only command ask for confirmation as well. Team Lead Jung detects and recovers from its own unresponsive state, cutting downtime from 2-3 minutes to 5 seconds, and a failed command now returns the error text and exit code so the next step can be decided.

Right-click menu order on nodes and groups

Node list

The menu runs read, create, organisation, agent, power, so the riskier the action the lower it sits. Group menus follow the same order. Nodes with no assigned employee were being dropped entirely when the top-level group was right-clicked; that is fixed.

Consistent widget labels and a faster home screen

Home dashboard

The same metric used to appear in a different colour in each widget; CPU, memory, I/O and handles now have fixed colours and timestamps share one format. Opening was reworked too, bringing the first screen from 3.6 seconds to 1.0.

Process record retention and server resources

Server

Raw process records are now deleted automatically after 180 days. The period can be tuned to the size of the site, down to a floor of 56 days. Until now these records could not be deleted at all. Network device polling that held hundreds of megabytes on every check was also fixed.

Agent clock offset reading nine hours out

Agent

The clock offset now shows its real value. The time the server sends is UTC but was not labelled as such, so the receiving side read it as local time and converted it again.