InnerWeb
Skip to main content

Update window without shift downtime: implementation notes

A night shift at the plant works with fewer people, so an update of the safety system at the wrong moment is most risky for it. Below are implementation notes: how to choose the window, what order to follow, which pitfalls to avoid, and which tests must pass before the operator returns to work.

2-2.5 h
full run of browser tests
Source: product documentation, E2E browser tests

Starting rule: an update must not take away an alarm

A work safety system has one function that must not be lost: the alarm. Before the plant team schedules a window, it should answer the question of what happens to alarms during and after the update. The answer must be concrete: who will receive an SOS signal and how, when the system is unavailable or restarting.

Good practice is a backup plan for the window: a voice or radio procedure, a designated person, and the length of the window written into the shift schedule. The vendor provides an estimate of the time, and the plant decides the moment.

Step 1: choosing the moment

We choose the moment to fit the plant's rhythm, not the vendor's convenience. The best window is one in which activity in the system is lowest and the IT and safety teams are available for the first checks. For a plant working continuously this often means a shift change or a technological break, rather than the middle of the night with a single on-call person.

The full set of browser tests takes about 2 to 2.5 hours and runs on request and automatically at midnight on weekends. This is a practical hint: a fresh result of a full run is produced over the weekend, so an update planned for the start of the week can rely on the result from the last weekend night.

Duration of a full test run against the window length

Duration of a full test run against the window lengthBar chart: a full run of the tests takes from 120 to 150 minutes.050100150minFull run, lower limit: 120 min120 minFull run, lower limitFull run, upper limit: 150 min150 minFull run, upper limit
Source: Product documentation: E2E browser tests; 2 hours = 120 min, 2.5 hours = 150 min

Step 2: the order of work

The order we recommend is simple. First, confirm the result of the regression tests from the last run. Then a data backup and an agreed version to return to if needed. Next the update, and right after it a check of the critical scenarios: logging in, work permits, the map, the alarm. Only after this check is the change released to normal work.

The test database is rebuilt from migrations once per test run, so the test result does not depend on data from the previous run. In practice, a check in the test environment leaves no traces that would distort the next check.

Pitfall: errors after an update that are not visible

After an update there is a risk of hidden errors that show no message but silently stop one function with an uncaught exception. That is why error monitoring should be active during the window, and exceptions describing invalid input, such as lack of access or a malformed request, are excluded from reporting as non-failures. The on-call person then sees the real errors without the noise.

Pitfall: alarms muted when the cache fails

Work safety alarms are not muted by a cache failure: if the cache fails, the alarm still rings and the log records the problem. This is a rule worth confirming in a test after the update, rather than assuming it.

Step 3: checks after the update

The check after the update should be short and repeatable. We recommend a list of five scenarios carried out by a named person at the plant: logging in, opening a work permit, opening the map, triggering a test alarm and checking the version. Each scenario has a pass criterion written down before the window.

The result is recorded in the window protocol, together with the start and end time and the person who released the change to work. If one scenario fails, the team decides: a fix on site or a return to the previous version. The condition is that the decision is taken before the end of the window, not after the operator has returned to work.

Conditions for success and integrations

InnerWeb is a software producer and integrates with the client's existing systems, such as access control, time and attendance, HR and ERP. An update on the InnerWeb side does not replace the readers or cards of the plant, so the window concerns the software layer. Still, establish whether the client's system will also be changed in the same window: two changes at once make it harder to find the cause when something does not work.

The last condition is the role of a person. Someone at the plant must decide to start the window and to release the shift to work. Without such a person the window turns into an operation of the vendor rather than a shared decision. Record the role by name in the procedure, not in this text.

  • a backup plan for receiving alarms during the window
  • a window matched to the shift rhythm, not to the vendor's convenience
  • the result of the regression tests from the last full run
  • a data backup and a known version to return to
  • five check scenarios after the update with pass criteria
  • a window protocol with the time and the person releasing the change

Video

Frequently asked questions

Can the full test set be run right after the update?

A full set takes about 2 to 2.5 hours, so during the window a short list of critical scenarios is normally used. The full result comes from the last run, for example the weekend one.

Does a software update replace readers or cards?

No. InnerWeb is a software producer and integrates with the client's existing systems, so the window concerns the software layer.

What about alarms if the cache fails?

Work safety alarms are not muted: if the cache fails, the alarm still rings and the log records the problem.

Let's plan the update window together

Together we will choose the moment, the order and the list of check scenarios so that a night shift does not lose its alarms. Book a talk.

Let's talk