code wiki / _hdl_build / nx_pub_recover.nx

nx_pub_recover.nx

buildroot/runtime/_hdl_build/nx_pub_recover.nx

7809 B140 linesdepth 3pulls 4 transitivereach 20 importersview sourcekind tooltopic pub

docs dependencies structs consts functions

about

nx_pub_recover.nx -- THE PUBLISHER'S DAEMON-RECOVERY rung (ops authority, not ad-hoc workstreams). Operator law: "the nishi publisher should be managing restarts and all this stuff." So recovery is a PUBLISHER responsibility, decided from the MONITOR'S signal -- never an ad-hoc workstream kill. The supervisor (nx_hostctl) already RESPAWNS dead/hung daemons + crash-loop-guards. The GAP it can't see: a WEDGE -- a daemon that still answers but SLOWLY (the :8791 reader at 20s/request passes the binary up/down probe). nx_apm_sweep measures real latency -> apm_metrics.tsv. This organ reads that signal, maps a slow endpoint to its daemon, and emits an IDEMPOTENT + SERIALIZED + LEDGERED recovery REQUEST (the control-plane consumer proc_kill_by_name's it -> the existing keeper respawns a fresh one). By construction: never-brick #26 (a recovery REQUEST is reversible metadata; the action is only a kill+supervised-respawn, never a destructive write); idempotent #10 (a daemon already PENDING is not re-requested); serialized (fl_acquire, one decision at a time); fail-safe #14 (unmapped/healthy -> no-op). pr_recover(metricspath, queuepath, wedge_ms) -> number of recovery requests emitted license_tier: ORIGINAL

dependencies 3 imports · 3 importers

imports: nx_syscalls.nxnx_arbiter.nxnx_framed_append.nx

imported by: nx_hostop.nxnx_pub_recover_exec.nxnx_pub_recover_gate.nx

call flow from main pre-order; caps 40 nodes / depth 6 declared; ↻ = already shown

structs

none

consts

20	const PR_MAGIC_100000: i64 = 100000
22	const PR_WEDGE_MS_DEFAULT: i64 = 5000
23	const PR_METRICS: *u8 = "knowledge/status/apm_metrics.tsv"
24	const PR_QUEUE: *u8 = "knowledge/publish/recovery_queue.tsv"
25	const PR_LOCK: *u8 = "publish_recover"
26	const PR_REC_CAP: i64 = 1024

functions

28	func pr_slen(s: *u8) -> i64 { var n: i64 = 0; while s[n] != (0 as u8) { n = n + 1 } return n } called by 3: pr_uw pr_has pr_recover
29	func pr_uw(fd: i64, s: *u8) -> i64 { sys_write(fd, s, pr_slen(s)); return 0 } called by 1: main calls 2: sys_write pr_slen
30	func pr_n(buf: *u8, off: i64, v: i64) -> i64 called by 6: pr_wn pr_recover pe_exec main gt_wn gt_row calls 1: sys_mmap
39	func pr_s(buf: u8, off: i64, s: u8) -> i64 { var o: i64 = off; var i: i64 = 0; while s[i] != (0 as u8) { buf[o] = s[i]; o = o + 1; i = i + 1 } return o } called by 10: pr_daemon_for pr_recover pe_exec main gt_row pp_decide+4
40	func pr_wn(fd: i64, v: i64) -> i64 { let b: *u8 = sys_mmap(24); let e: i64 = pr_n(b, 0, v); sys_write(fd, b, e); return 0 } called by 1: main calls 3: sys_mmap pr_n sys_write
42	func pr_has(hay: u8, n: i64, needle: u8) -> i64 called by 1: pr_daemon_for calls 1: pr_slen
49	func pr_field(line: u8, linelen: i64, idx: i64, out: u8) -> i64 called by 6: pr_already_pending pr_recover pe_exec pp_recent_recoveries pp_map_lookup pp_decide
59	func pr_atoi(s: u8) -> i64 { var v: i64 = 0; var i: i64 = 0; while s[i] != (0 as u8) { let c: i64 = s[i] as i64; if c >= 48 { if c <= 57 { v = v10 + (c-48) } } i = i + 1 } return v } called by 2: pr_recover main
61	func pr_streq(a: u8, b: u8) -> i64 { var i: i64 = 0; while a[i] != (0 as u8) { if a[i] != b[i] { return 0 } i = i + 1 } if b[i] != (0 as u8) { return 0 } return 1 } called by 1: pr_already_pending
64	func pr_daemon_for(url: u8, urllen: i64, out: u8) -> i64 called by 1: pr_recover calls 2: pr_has pr_s
73	func pr_already_pending(queuepath: u8, daemon: u8) -> i64 called by 2: pr_recover pp_decide calls 4: sys_mmap sys_read_file pr_field pr_streq
94	func pr_recover(metricspath: u8, queuepath: u8, wedge_ms: i64) -> i64 called by 3: hostop_recover main main calls 13: sys_mmap sys_read_file pr_field pr_atoi pr_slen pr_daemon_for+7
132	func main(argc: i64, argv: *i64) -> i64 calls 4: pr_atoi pr_uw pr_wn pr_recover