KVS connection monitor
pod kvsmon-6d87d75c44-c9g25 · egress IP 5.148.184.24 · running 4660s
· 2026-08-05T16:35:47Z
targets: public → rediss://kvsmon-redis.00e40d6.keyvaluestore.nineapis.ch:6379
SVC-114637 — looking for idle-flow resets on the deplo.io → KVS path.
Each row is one dedicated long-lived connection. idle@drop is how long the flow had been
quiet when it was killed — that number is the intermediary's idle timeout.
Probes
| probe | state | doing | up | idle | drops |
conn.fail | idle@drop min | med | max | longest life |
peer | last error |
|---|
| public/stream-block | up | blocked_read | 4660.1s | 160.0s | 0 | 0 | – | – | – | – | 178.209.60.11:6379 | – |
| public/stream-block-ka | up | blocked_read | 4660.1s | 160.0s | 0 | 0 | – | – | – | – | 178.209.60.11:6379 | – |
| public/ping-60 | up | idle | 4660.1s | 39.9s | 0 | 0 | – | – | – | – | 178.209.60.11:6379 | – |
| public/ping-180 | up | idle | 4660.1s | 160.0s | 0 | 0 | – | – | – | – | 178.209.60.11:6379 | – |
| public/ping-300 | up | idle | 4660.1s | 160.0s | 0 | 0 | – | – | – | – | 178.209.60.11:6379 | – |
| public/busy | up | busy | 4660.1s | 0.6s | 0 | 0 | – | – | – | – | 178.209.60.11:6379 | – |
| public/busy-ping | up | busy | 4660.1s | 0.3s | 0 | 0 | – | – | – | – | 178.209.60.11:6379 | – |
public/stream-block: XREADGROUP BLOCK 900s, no TCP keepalive — same shape as Symfony Messenger · public/stream-block-ka: same, but SO_KEEPALIVE with 30s idle — tests whether client keepalive prevents the drop · public/ping-60: idle 60s between PINGs — drops here bracket the intermediary's idle timeout · public/ping-180: idle 180s between PINGs — drops here bracket the intermediary's idle timeout · public/ping-300: idle 300s between PINGs — drops here bracket the intermediary's idle timeout · public/busy: 1024B SET+GET every 1s, never idle — the counter-test: if this drops too, the cause is not an idle timeout and the search moves elsewhere · public/busy-ping: PING every 1s — active but carrying no data. Paired with `busy` it separates 'the flow is active' from 'the flow is moving data'
Incidents (newest first, last 100)
| time (UTC) | probe | event | during | idle | alive |
error class | error |
|---|
| no drops recorded yet |
This page shows this pod only and is lost on restart. The durable record across
all replicas is on stdout, one JSON line per event:
nctl logs app kvsmon -p josi --follow