Reconstructing a silent outage from counters you already keep
A silent outage, in the sense we use the phrase at Releasegridpath, is a period when a sector stops carrying successful speech or data sessions while the alarm list the night desk watches does not light. The counters still move. CSSR on that sector falls, attempts fall, or both. Neighbouring sectors pick up a fraction of the load and look busy rather than failed.
Start with the site code as it exists in the counter file, not the marketing name on the complaint form. Marketing names drift; site codes in the extract usually do not. Write the code at the top of the note before you read anything else. If two codes exist for the same cabinet after a swap, record both and say which file used which.
Pull hourly attempts and successes for the sector and for its first-tier neighbours for the same clock. A genuine silent failure often shows attempts collapsing on the failing sector while neighbours rise. A transmission bottleneck further up can show all of them falling together. That distinction is the first fork in the chronology.
Do not average the day. A two-hour hole at 02:00 disappears inside a daily mean and then reappears in the morning as a vague ‘quality dip’. Keep the hour. If the extract is 15-minute, keep 15 minutes. The walkthrough note should quote the same grain the file uses.
Tickets written during the window are evidence only if their timestamps can be compared to the counter clock. If the ticket system and the OSS sit in different zones, write both. We have seen a ‘three-hour outage’ that was a one-hour failure plus a two-hour offset. The counters were not wrong; the clocks were uncompared.
Plant notes belong in the same chronology: generator run hours, fibre tickets from the transmission group, landlord power cuts. They do not become the cause until a timestamp lines up. If the generator log starts after the counter collapse, say so. A gap is more useful than a tidy story.
Releasegridpath’s local utility can hold this chronology on the desk’s own machine. It does not subscribe to the OSS. Someone still has to place the extract in the folder. That limitation is intentional: the record stays where the operator already keeps it.