InnerWeb
Skip to main content

Server room alarm on a plant network without outbound traffic

A plant network often blocks all outbound traffic from the application host, yet an alarm about an unauthorised entry to the server room still has to reach the on-duty phone. Below are notes for the network administrator: how the durable webhook queue works, where the poller sits and what to check before going live.

0
outbound rules to open for the host
Source: feature card
1
durable queue for alarms
Source: feature card

The problem: an alarm that cannot be sent out

A network administrator in a manufacturing plant closes outbound traffic from the server room segment, because the security policy requires it. The application host cannot trigger an external webhook on its own or send a notification to a service outside the network. At the same time, an alarm about an unauthorised entry cannot be lost, because then it has no meaning.

The typical reaction is to open outbound traffic to a specific address. This may be unacceptable, because it requires policy changes, approval from the security department and testing. The alternative is simpler for the administrator: open nothing, and only reverse the direction of the data flow.

The solution: a durable queue and an external poller

When the client's network blocks all outbound traffic from the application host, alarm webhooks are written to a durable queue, which is collected by an external poller. The alarm is not lost without opening outbound traffic. The host does not initiate a connection to the outside: the poller on the external side asks the queue and collects the events.

From the administrator's point of view this means one inbound rule for the poller instead of an outbound rule for the host. The poller runs in a segment that is allowed to reach the on-duty phone, a mobile network or a notification system. The queue is durable, so a host restart does not wipe pending alarms.

This setup requires planning the polling frequency. Polling too rarely delays the alarm, polling too often loads the host. The value is set together with the on-duty person: how many seconds of delay are acceptable for an alarm about an entry to the server room.

Where an alarm about an unauthorised entry comes from

The alarm is generated by an automation rule built in a no-code creator. The rule reacts to a change in the permit status, to an alarm, to a schedule or to the presence of an employee, and sends an email, push, webhook or opens a permit from a template. For the server room the typical rule is: a rejected badge event at the server room door or an entry without a permit, and then a webhook to the on-duty person.

The source of badge events is the client's KD system. InnerWeb receives card badge events through a webhook and links them with qualifications, permits, attendance and time and attendance (RCP) data. It does not manufacture or replace access control readers. The alarm therefore rests on events that already exist in the plant.

It is good to separate the levels. An entry without a permit within the window is an alarm for the on-duty IT person. A rejected badge is an informational notice. One rule for everything floods the on-duty person and leads to ignored notifications.

Conditions before go-live and tests

Before the start, check four things. First, where the poller sits and what connection it has to the on-duty phone. Second, how long the queue keeps events when the poller does not respond. Third, who is on duty outside working hours. Fourth, what happens when the on-duty person does not confirm the alarm.

Acceptance test: trigger a controlled alarm with a test card at the door and measure the time from the badge to the notification on the phone. Repeat the test after a host restart to check the durability of the queue. Record the result as acceptance evidence, because for an audit that is more valuable than a declaration.

Multi-factor authentication on accounts that manage alarm rules reduces the risk that someone switches an alarm off. Changes to automation rules are saved together with their author and time, so disabling a rule leaves a trace.

Pitfalls and good practices

The first pitfall: a poller without monitoring. If it fails, alarms will quietly wait in the queue and nobody will know. Add a simple heartbeat signal for the poller. The second: no procedure for the on-duty person. A notification without knowing what to do is just noise. The third: testing only during working hours.

Good practice: a decision map on one page. Who receives the alarm, within how many minutes they must react, whom they inform next and how they close the event. A closed event is recorded in the log together with its justification, so the next audit sees the full history.

Checklist

Video

Frequently asked questions

Do alarms require opening outbound traffic?

No. Alarm webhooks are stored in a durable queue, and an external poller collects them. The host does not need to initiate connections to the outside.

What if the poller stops working?

Events wait in the queue, so they are not lost, but the alarm is delayed. That is why the poller requires monitoring and a heartbeat signal.

Does the alarm depend on our KD system?

Yes, the source of badge events is the client's KD system, which passes events through a webhook. InnerWeb does not manufacture or replace readers.

Let's talk about evidence of server room entries

We will show in a demo how badge events from your access control system connect with the permit and the log. No reader replacement and no card changes.

Let's talk