accesslog-shipper: emit __shipstat__ sent-line checkpoints for splits reconciliation

The splits completeness floor is TIME-based (trust hours complete since shipper
connect) and blind to event-loss-by-rotation (2026-07-10: raw_events rotated
every 1-2min, cruncher lost ~85% of us-32 events, floor still called the hours
complete). Shipper now injects {"__shipstat__":1,"sess":..,"sent":N,"ts":..}
every 1000 lines or 30s through the SAME tcp pipe. Cruncher diffs shipper-sent
vs DB-ingested per gateway/hour -> a reconciliation-based floor that catches the
loss a clock cannot. sess=connection-start epoch so resets on reconnect are
detectable. Logic in a mounted .awk file (not inline) to dodge sh/awk quoting.
Verified on busybox:1.36 (systime/fflush OK). NOT deployed until claude@'s
cruncher filters __shipstat__ from event parsing - coordinating via SIP.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
rob
2026-07-12 02:40:58 +00:00
parent ff41ec1c25
commit 932617814a
2 changed files with 24 additions and 1 deletions

18
accesslog-shipper.awk Normal file
View File

@@ -0,0 +1,18 @@
# Streams dshackle access-log lines unchanged, and every 1000 lines OR 30s
# injects a __shipstat__ checkpoint carrying the cumulative sent-count for this
# connection. The splits cruncher (claude host) filters these OUT of event
# processing and uses them to reconcile shipped-vs-ingested per gateway per hour
# — the completeness signal the time-based floor cannot see (event-loss-by-
# rotation, 2026-07-10). `sess` = connection-start epoch so the cruncher detects
# counter resets on shipper reconnect (retry loop resumes at file END).
BEGIN { sess = systime(); last = sess; n = 0 }
{
n++
print
now = systime()
if (n % 1000 == 0 || now - last >= 30) {
printf "{\"__shipstat__\":1,\"sess\":%d,\"sent\":%d,\"ts\":%d}\n", sess, n, now
fflush()
last = now
}
}

View File

@@ -11,9 +11,14 @@ services:
restart: unless-stopped restart: unless-stopped
volumes: volumes:
- ./dshackle_logs:/logs:ro - ./dshackle_logs:/logs:ro
# sent-line counter: injects __shipstat__ checkpoints into the stream so
# the listener cruncher can reconcile shipped-vs-ingested per gateway/hour
# (fixes the time-based completeness floor being blind to event-loss-by-
# rotation, 2026-07-10). Cruncher MUST filter __shipstat__ from events.
- ./accesslog-shipper.awk:/shipper.awk:ro
command: >- command: >-
sh -c 'while true; do sh -c 'while true; do
tail -F /logs/access_log.jsonl 2>/dev/null | nc ${SPLITS_LISTENER_HOST:-78.47.193.225} ${SPLITS_LISTENER_PORT:-9102}; tail -F /logs/access_log.jsonl 2>/dev/null | awk -f /shipper.awk | nc ${SPLITS_LISTENER_HOST:-78.47.193.225} ${SPLITS_LISTENER_PORT:-9102};
echo shipper lost listener connection, retrying in 10s; echo shipper lost listener connection, retrying in 10s;
sleep 10; done' sleep 10; done'