diff options
Diffstat (limited to 'prompts/skills/f3s-storage/references/zrepl.md')
| -rw-r--r-- | prompts/skills/f3s-storage/references/zrepl.md | 32 |
1 files changed, 20 insertions, 12 deletions
diff --git a/prompts/skills/f3s-storage/references/zrepl.md b/prompts/skills/f3s-storage/references/zrepl.md index ea31644..7c17be1 100644 --- a/prompts/skills/f3s-storage/references/zrepl.md +++ b/prompts/skills/f3s-storage/references/zrepl.md @@ -201,31 +201,39 @@ doas zfs set readonly=on zdata/sink/f0/zdata/enc/nfsdata # prevent replication ## Replication liveness check A small cron job on f0 writes a canary file into the replicated dataset so its -presence and mtime on f1's read-only sink reveals how fresh the zrepl -replication is. +content and mtime on f1's read-only sink reveal how fresh the zrepl replication +is. On **f0** (root crontab), every 10 minutes, gated on a sentinel file so the -canary only runs when the dataset is actually in use: +canary only runs when the dataset is actually in use, the job touches the file +and writes the current unix epoch into it: ``` -*/10 * * * * test -f /data/nfs/nfs.DO_NOT_REMOVE && /usr/bin/touch /data/nfs/nfs.LIVE_CHECK +*/10 * * * * test -f /data/nfs/nfs.DO_NOT_REMOVE && /usr/bin/touch /data/nfs/nfs.LIVE_CHECK && /bin/date +\%s > /data/nfs/nfs.LIVE_CHECK ``` +Note: the `\%` is required — cron treats an unescaped `%` as a newline/stdin +separator, so `date +%s` would silently break. In an interactive shell you'd +write `date +%s` (no backslash). + `/data/nfs` on f0 is `zdata/enc/nfsdata` — the dataset the `f0_to_f1_nfsdata` -job pushes to f1 every minute. zrepl snapshots + sends, so the touched -`nfs.LIVE_CHECK` (and its mtime) appear on f1's read-only sink at +job pushes to f1 every minute. zrepl snapshots + sends, so `nfs.LIVE_CHECK` +(its epoch content and mtime) appears on f1's read-only sink at `/data/nfs/nfs.LIVE_CHECK` within roughly one replication interval. -On **f1**, read the replica to check freshness (the sink is read-only): +On **f1**, read the replica to check freshness (the sink is read-only) — either +compare the epoch inside the file to f1's clock, or compare mtimes: ```sh -stat -f '%Sm' /data/nfs/nfs.LIVE_CHECK # f0's last touch time, as replicated -date '+%Sm' # compare to now +echo $(( $(date +%s) - $(cat /data/nfs/nfs.LIVE_CHECK) )) # seconds since f0 last wrote (expect <~660) +stat -f '%Sm' /data/nfs/nfs.LIVE_CHECK # f0's last write time, as replicated +date '+%Sm' # compare to now ``` -If the mtime lags more than ~10–11 minutes behind real time, zrepl replication -is stalled (see Troubleshooting). The `/data/nfs/nfs.DO_NOT_REMOVE` sentinel -also doubles as the zrepl-replication canary guard. Installed on f0 2026-07-20. +If the lag exceeds ~10–11 minutes (10-min cron interval + 1-min zrepl +replication), zrepl replication is stalled (see Troubleshooting). The +`/data/nfs/nfs.DO_NOT_REMOVE` sentinel also doubles as the canary guard. +Installed on f0 2026-07-20. ## Failover design: intentionally read-only replica |