One day on the pager

A Tuesday with a bad night in it, and every control the app has, in the order it came up.

Nothing below is unusual. It is one ordinary day of one person carrying one phone, and it happens to use every field a send takes and every control the phone has over what arrives. The times are local; the sends are the real ones.

the six channels this day runs on

None of them was made that morning. A channel is a name, an address and a short list of defaults: how loud it is allowed to get, what it does at night, and who it rings if nobody answers. The address is the whole integration. Whatever holds it can post to that channel and to nothing else, so the deploy script and the database monitor never have to know about each other.

Four of the six are ordinary channels, sitting there waiting to be written to. The other two are vitals, which work the other way round: nothing posts to say something broke, something posts to say it is still alive, and the alert is the beat that never comes. Both kinds share one list on the phone. The vitals go at the top as a wall of dots, because there is only ever one thing to know about each.

Nothing in the commands below chooses a sound, a repeat, an escalation partner or a quiet window. All of that lives on the channel and is set once. A send asks for a level and passes the text. priority=page asks for the top of the channel’s range, and the highest priority in the table above decides whether it gets there.

$ curl -X POST https://api.tissue.systems/v1/mast/channels/mc_2b7d…/mute \
    -H "Authorization: Bearer $TISSUE_TOKEN" \
    -d '{"until": "2026-09-15T14:22:00Z"}'

$ curl -d silent=1 -d body="deployed 2026.09.029 · s10, s33" "$MAST_URL"

$ curl -d priority=loud \
    -d title="nightly build failed" \
    -d url="https://github.com/…/actions/runs/35055965614" \
    -d url_title="Open the run" "$MAST_URL"

$ curl -d key=queue-consumer -d priority=loud \
    -d title="consumer cannot reach redis" "$MAST_URL"
{"ok":true,"state":"deduped","duplicate":true}

$ curl -d timestamp=2026-09-15T17:10:00Z \
    -d title="import finished" \
    -d body="18,204 rows · 2 rejected" "$MAST_URL"

# the app → Pause pages → 1 hour
# banner: pages paused · until 14:00
# still rings: Home · smoke

# 202 {"ok":true,"state":"queued"}
# the feed: held · pages paused until 14:00

# the card: asked for page · held 22 min

$ curl -d priority=page -d key=disk-var \
    -d retry=600 -d expire=86400 \
    -d title="/var 91% full" "$MAST_URL"

$ curl -d key=disk-var -d resolve=1 \
    -d title="/var back to 61%" "$MAST_URL"
{"ok":true,"state":"resolved","open_for":5700}

# the app → the channel → Quiet hours → 19:00–07:00

# one notification: 34 messages · api-health
# answers come back {"state":"suppressed"}

$ curl -d priority=page -d key=repl-lag \
    -d sound=klaxon -d title="db-01" \
    -d body="no WAL segment in 11 minutes" "$MAST_URL"

# 03:04 acked · 6 min after fired
# still open

# 03:31 escalated · partner
# nothing on the send says this — escalation
# is a property of the channel, set in the app

17 3 * * *  /usr/local/bin/backup.sh \
    && curl -fsS https://mast.tissue.dev/mk_8e2a…

# 07:20 the crontab line runs
# 202 {"ok":true,"state":"alive"} — no card, so no id
# the open page: resolved · open for 1 h 20 m

# fired → escalated → acked → resolved
# with the time of each

All eighteen land in the same list, in the order they happened, including the ones that made no sound.

where each control appeared

So that nothing is hiding in the story.

Two of the eighteen moments are a person answering, and one is a job answering on its own. The rest of the day the phone is silent, which is where a pager sits almost all of the time.

The overview and the price are on the Tissue Mast page, and everything else is in the docs.