diff --git a/files/etc/init.d/owrt-remote-logs b/files/etc/init.d/owrt-remote-logs index 4aa707b..eaec441 100644 --- a/files/etc/init.d/owrt-remote-logs +++ b/files/etc/init.d/owrt-remote-logs @@ -8,7 +8,8 @@ start_service() { /bin/sh /usr/lib/owrt-remote/logs.sh migrate || logger -t owrt-remote 'WARN: старый RAM-журнал ещё не перенесён; см. свободное место Flash.' procd_open_instance journal procd_set_param command /bin/sh /usr/lib/owrt-remote/logs.sh collect - procd_set_param respawn 3600 5 5 + # Retry at a bounded rate even after repeated logd/storage outages. + procd_set_param respawn 3600 10 0 procd_set_param stdout 1 procd_set_param stderr 1 procd_close_instance diff --git a/files/usr/lib/owrt-remote/logs.sh b/files/usr/lib/owrt-remote/logs.sh index dfc8e52..11f47b8 100644 --- a/files/usr/lib/owrt-remote/logs.sh +++ b/files/usr/lib/owrt-remote/logs.sh @@ -7,10 +7,20 @@ LOG_PART_LIMIT=256 LOG_BUFFER_LIMIT=16384 LOG_BATCH_LIMIT=16384 LOG_FLUSH_SECONDS=60 +LOG_HEARTBEAT_SECONDS=300 LOG_FREE_KB=2048 log_clock() { cut -d. -f1 /proc/uptime; } log_epoch() { date +%s; } +log_pause() { + # Some OpenWrt BusyBox builds omit fractional sleep. Keep lock/reap deadlines + # in tenths, including the one-second fallback, instead of spinning on failure. + if [ "${LOG_PAUSE_TENTHS:-0}" != 10 ]; then + if sleep 0.1 2>/dev/null; then LOG_PAUSE_TENTHS=1; return 0; fi + LOG_PAUSE_TENTHS=10 + fi + sleep 1 +} log_self_pid() { # $$ is the parent shell's PID inside ash subshells/pipelines. Reading proc # with the shell builtin yields the actual lock holder, including the writer. @@ -105,8 +115,8 @@ log_lock() { [ "$attempt" -lt 20 ] || log_reap_lock "$RUN_DIR/lock" "$attempt" ;; *) kill -0 "$owner" 2>/dev/null || log_reap_lock "$RUN_DIR/lock" "$attempt" ;; esac - attempt=$((attempt + 1)); [ "$attempt" -lt 100 ] || return 1 - sleep 0.1 + log_pause + attempt=$((attempt + LOG_PAUSE_TENTHS)); [ "$attempt" -lt 100 ] || { log_diag lock-timeout; return 1; } done log_self_pid >"$RUN_DIR/lock/pid" # Another first writer may have selected storage while we waited. @@ -139,8 +149,24 @@ log_reap_lock() { owner="$(cat "$1/pid" 2>/dev/null)" case "$owner" in ''|*[!0-9]*) rm -f "$1/pid" ;; - *) if kill -0 "$owner" 2>/dev/null; then rm -f "$1/reap/pid"; rmdir "$1/reap"; return 0; else rm -f "$1/pid"; fi ;; + *) + if { [ "$1" = "$RUN_DIR/collector-lock" ] && log_process_info "$owner" >/dev/null; } || + { [ "$1" != "$RUN_DIR/collector-lock" ] && kill -0 "$owner" 2>/dev/null; }; then + rm -f "$1/reap/pid"; rmdir "$1/reap"; return 0 + else rm -f "$1/pid"; fi ;; esac + if [ "$1" = "$RUN_DIR/collector-lock" ] && [ -s "$1/reader" ]; then + local old_reader old_birth + read -r old_reader old_birth <"$1/reader" + if [ -n "$old_birth" ] && [ "$(log_process_info "$old_reader")" = "$old_birth" ]; then + kill "$old_reader" 2>/dev/null || true + local remaining + remaining=0 + while [ "$(log_process_info "$old_reader")" = "$old_birth" ] && [ "$remaining" -lt 10 ]; do log_pause; remaining=$((remaining + LOG_PAUSE_TENTHS)); done + if [ "$(log_process_info "$old_reader")" = "$old_birth" ]; then kill -KILL "$old_reader" 2>/dev/null || true; fi + fi + rm -f "$1/reader" + fi rm -f "$1/reap/pid"; rmdir "$1/reap"; rmdir "$1" 2>/dev/null || true } log_fallback_locked() { @@ -397,6 +423,7 @@ log_flush_locked() { TOTAL=$((TOTAL + bytes)) log_checkpoint_locked && log_commit_locked || { log_warn_locked write-warning 'Flash: ошибка фиксации состояния журнала.'; return 1; } : >"$RUN_DIR/pending" + printf '%s\n' "$(log_epoch)" >"$RUN_DIR/commit-health" rm -f "$RUN_DIR/inflight" rm -f "$RUN_DIR/write-warning" "$RUN_DIR/queue-warning" "$RUN_DIR/buffer-warning" dropped="$(cat "$RUN_DIR/dropped" 2>/dev/null)"; dropped="${dropped:-0}" @@ -406,6 +433,7 @@ log_flush_locked() { fi fi log_clock >"$RUN_DIR/last-flush" + printf '%s %s\n' "$(log_epoch)" "$(cat "$RUN_DIR/last-flush")" >"$RUN_DIR/flush-health.new" && mv -f "$RUN_DIR/flush-health.new" "$RUN_DIR/flush-health" } log_open() { log_lock || return 1 @@ -439,14 +467,22 @@ log_redact() { }' } log_new_fact_locked() { - local key now heartbeat new_fact + local key now heartbeat new_fact last # Compare event text without syslog timestamp, facility, PID or Xray time. key="$(printf '%s\n' "$1" | tr '\t' ' ' | sed -e 's/^\[[^]]*\] *//' -e 's/.*owrt-remote\(\[[0-9][0-9]*\]\)\{0,1\}: *//' -e 's/^[0-9][0-9]*\/[0-9/:. ]* *//' -e 's/ *(\([0-9][0-9]*\)\/[0-9][0-9]*)//')" heartbeat=0 case "$key" in 'Heartbeat: OK;'*) heartbeat=1 - if [ "$(cat "$RUN_DIR/heartbeat-ok" 2>/dev/null)" = "$key" ]; then return 1; fi + last=0 + [ ! -f "$RUN_DIR/heartbeat-stored" ] || read -r last <"$RUN_DIR/heartbeat-stored" + case "$last" in ''|*[!0-9]*) last=0 ;; esac + now="$(log_clock)" + # Keep one liveness sample every five minutes, plus every change/recovery. + # Input freshness is tracked in RAM even when a duplicate is suppressed. + if [ "$(cat "$RUN_DIR/heartbeat-ok" 2>/dev/null)" = "$key" ] && + [ "$now" -ge "$last" ] && [ "$((now - last))" -lt "$LOG_HEARTBEAT_SECONDS" ]; then return 1; fi + printf '%s\n' "$now" >"$RUN_DIR/heartbeat-stored" printf '%s\n' "$key" >"$RUN_DIR/heartbeat-ok" # A recovery lets the same failure appear again; keys are fixed-size hashes. : >"$RUN_DIR/facts" ;; @@ -479,24 +515,50 @@ log_new_fact_locked() { [ "$new_fact" = 0 ] || return 1 return 0 } +log_event_health_locked() { + local epoch uptime accepted_epoch accepted_up accepted suppressed failed extra + accepted_epoch=0; accepted_up=0; accepted=0; suppressed=0; failed=0 + if [ -s "$RUN_DIR/event-health" ]; then + read -r epoch uptime accepted_epoch accepted_up accepted suppressed failed extra <"$RUN_DIR/event-health" + case "$accepted_epoch:$accepted_up:$accepted:$suppressed:$failed" in *[!0-9:]*|:*|*:) accepted_epoch=0; accepted_up=0; accepted=0; suppressed=0; failed=0 ;; esac + fi + epoch="$(log_epoch)"; uptime="$(log_clock)" + case "$1" in + written) accepted_epoch="$epoch"; accepted_up="$uptime"; [ "$accepted" -ge 2147483647 ] || accepted=$((accepted + 1)) ;; + suppressed) [ "$suppressed" -ge 2147483647 ] || suppressed=$((suppressed + 1)) ;; + failed) [ "$failed" -ge 2147483647 ] || failed=$((failed + 1)) ;; + esac + printf '%s %s %s %s %s %s %s\n' "$epoch" "$uptime" "$accepted_epoch" "$accepted_up" "$accepted" "$suppressed" "$failed" >"$RUN_DIR/event-health.new" && + mv -f "$RUN_DIR/event-health.new" "$RUN_DIR/event-health" +} log_append() { - local line size bytes dropped + local line size bytes dropped outcome log_redact "${1:-OK}" | while IFS= read -r line; do - log_open || exit 1 - log_flush_locked || true + log_open || { log_diag append-open-failed; exit 1; } + if ! log_flush_locked; then log_diag flush-failed; fi [ -f "$RUN_DIR/facts" ] || : >"$RUN_DIR/facts" + outcome=suppressed if log_new_fact_locked "$line"; then size="$(wc -c <"$RUN_DIR/pending")"; bytes="$(printf '%s\n' "$line" | wc -c)" # Reserve room for a single overflow warning, too. if [ "$((size + bytes))" -le "$((LOG_BUFFER_LIMIT - 1024))" ]; then - printf '%s\n' "$line" >>"$RUN_DIR/pending" + if ! printf '%s\n' "$line" >>"$RUN_DIR/pending"; then + # A failed append must be eligible for retry, not hidden by dedup. + rm -f "$RUN_DIR/heartbeat-ok" "$RUN_DIR/facts" + log_event_health_locked failed; log_diag append-write-failed + log_close; exit 1 + fi + outcome=written else + outcome=failed + rm -f "$RUN_DIR/heartbeat-ok" "$RUN_DIR/facts" dropped="$(cat "$RUN_DIR/dropped" 2>/dev/null)"; dropped="${dropped:-0}" [ "$dropped" -ge 2147483647 ] || dropped=$((dropped + 1)) printf '%s\n' "$dropped" >"$RUN_DIR/dropped" log_warn_locked buffer-warning 'Буфер логов заполнен: новые события сверх лимита пропускаются, неподтверждённый журнал сохранён.' fi fi + log_event_health_locked "$outcome" || { log_diag health-write-failed; log_close; exit 1; } log_close done } @@ -542,7 +604,7 @@ log_reset_locked() { log_checkpoint_locked && log_commit_locked || return 1 rm -f "$LOG_DIR"/part.* "$LOG_DIR/damaged.log" : >"$RUN_DIR/pending" - rm -f "$RUN_DIR/ack" "$RUN_DIR/facts" "$RUN_DIR/heartbeat-ok" "$RUN_DIR/dropped" "$RUN_DIR/buffer-warning" "$RUN_DIR/inflight" + rm -f "$RUN_DIR/ack" "$RUN_DIR/facts" "$RUN_DIR/heartbeat-ok" "$RUN_DIR/heartbeat-stored" "$RUN_DIR/dropped" "$RUN_DIR/buffer-warning" "$RUN_DIR/inflight" log_clock >"$RUN_DIR/last-flush" } log_clear() { log_open || return 1; log_reset_locked; result=$?; log_close; return "$result"; } @@ -592,10 +654,16 @@ log_ack() { fi log_close } -log_flush() { log_open || return 1; log_flush_locked "${1:-}"; result=$?; log_close; return "$result"; } +log_flush() { + log_open || return 1 + log_flush_locked "${1:-}"; result=$? + [ "$result" = 0 ] || log_diag flush-failed + log_close; return "$result" +} log_status() { log_open || return 1 printf 'storage=%s\npersistent=%s\nruntime=%s\nlimit_bytes=%s\nmax_parts=%s\nbuffer_limit_bytes=%s\nflush_seconds=%s\njournal_id=%s\ncommitted_bytes=%s\nack_bytes=%s\npending_bytes=%s\n' "$LOG_DIR" "$PERSISTENT" "$RUN_DIR" "$LOG_LIMIT" "$LOG_PART_LIMIT" "$LOG_BUFFER_LIMIT" "$LOG_FLUSH_SECONDS" "$JOURNAL" "$TOTAL" "$ACK" "$(wc -c <"$RUN_DIR/pending")" + log_health_status log_close } log_migrate() { @@ -687,36 +755,183 @@ log_sync_loop() { sleep 3 done } +# Diagnostics contain fixed reason codes only, never a line, URL or credentials. +# At most 4 KiB plus a 2 KiB compaction file, all on tmpfs. No syslog feedback. +log_diag() { + local code now last previous bytes + code="$1"; now="$(log_clock)"; last=0; previous='' + [ -d "$RUN_DIR" ] || return 0 + mkdir "$RUN_DIR/diag-lock" 2>/dev/null || return 0 + if [ -s "$RUN_DIR/diag-last" ]; then read -r last previous <"$RUN_DIR/diag-last"; fi + case "$last" in ''|*[!0-9]*) last=0 ;; esac + if [ "$code" != "$previous" ] || [ "$now" -lt "$last" ] || [ "$((now - last))" -ge 60 ]; then + bytes=0; [ ! -f "$RUN_DIR/diagnostics" ] || bytes="$(wc -c <"$RUN_DIR/diagnostics")" + if [ "$bytes" -gt 3900 ]; then + tail -c 2048 "$RUN_DIR/diagnostics" >"$RUN_DIR/diagnostics.new" && mv -f "$RUN_DIR/diagnostics.new" "$RUN_DIR/diagnostics" + fi + printf '%s %s\n' "$(log_epoch)" "$code" >>"$RUN_DIR/diagnostics" + printf '%s %s\n' "$now" "$code" >"$RUN_DIR/diag-last" + fi + rmdir "$RUN_DIR/diag-lock" 2>/dev/null || true +} +log_process_info() { + local stat + case "$1" in ''|*[!0-9]*) return 1 ;; esac + kill -0 "$1" 2>/dev/null || return 1 + if ! read -r stat <"/proc/$1/stat" 2>/dev/null; then return 1; fi + stat="${stat##*) }" + set -- $stat + [ "$#" -ge 20 ] || return 1 + case "$1" in Z|X) return 1 ;; esac + shift 19 + printf '%s\n' "$1" +} +log_collector_lock() { + local owner attempt + attempt=0 + while ! mkdir "$RUN_DIR/collector-lock" 2>/dev/null; do + owner="$(cat "$RUN_DIR/collector-lock/pid" 2>/dev/null)" + case "$owner" in + ''|*[!0-9]*) [ "$attempt" -lt 20 ] || log_reap_lock "$RUN_DIR/collector-lock" "$attempt" ;; + *) + if log_process_info "$owner" >/dev/null; then return 2; fi + # Reap a logread orphan only if its PID AND kernel start time match. + log_reap_lock "$RUN_DIR/collector-lock" "$attempt" ;; + esac + log_pause + attempt=$((attempt + LOG_PAUSE_TENTHS)); [ "$attempt" -lt 40 ] || { log_diag collector-lock-timeout; return 1; } + done + log_self_pid >"$RUN_DIR/collector-lock/pid" +} +log_collector_progress() { + printf '%s %s %s %s %s %s\n' "$collector" "$(log_epoch)" "$(log_clock)" "$received" "$append_errors" "$flush_errors" >"$RUN_DIR/collector-progress.new" && + mv -f "$RUN_DIR/collector-progress.new" "$RUN_DIR/collector-progress" +} +log_collector_cleanup() { + local code owner attempt + code="$1"; trap - EXIT INT TERM + exec 3>&- + if [ -n "$reader" ]; then + kill "$reader" 2>/dev/null || true + # Never leave an uninterruptible wait in the procd shutdown path. + attempt=0 + while log_process_info "$reader" >/dev/null && [ "$attempt" -lt 10 ]; do log_pause; attempt=$((attempt + LOG_PAUSE_TENTHS)); done + if log_process_info "$reader" >/dev/null; then kill -KILL "$reader" 2>/dev/null || true; fi + wait "$reader" 2>/dev/null || true + fi + rm -f "$fifo" + owner="$(cat "$RUN_DIR/collector-lock/pid" 2>/dev/null)" + if [ "$owner" = "$collector" ]; then + # Flush while still owning the collector lock. The successor never overlaps. + (log_flush force) || true + rm -f "$RUN_DIR/collector-lock/reader" "$RUN_DIR/collector-lock/pid" + rmdir "$RUN_DIR/collector-lock" 2>/dev/null || true + fi + exit "$code" +} +log_health_status() { + local owner reader birth now state last_input input_up last_event event_up accepted suppressed failed extra flush_epoch flush_up received errors flush_errors progress_owner + owner=''; reader=''; birth=''; last_input=0; input_up=0; last_event=0; event_up=0; accepted=0; suppressed=0; failed=0; flush_epoch=0; flush_up=0; received=0; errors=0; flush_errors=0 + [ ! -s "$RUN_DIR/collector-lock/pid" ] || read -r owner <"$RUN_DIR/collector-lock/pid" + [ ! -s "$RUN_DIR/collector-lock/reader" ] || read -r reader birth <"$RUN_DIR/collector-lock/reader" + [ ! -s "$RUN_DIR/event-health" ] || read -r last_input input_up last_event event_up accepted suppressed failed extra <"$RUN_DIR/event-health" + [ ! -s "$RUN_DIR/flush-health" ] || read -r flush_epoch flush_up <"$RUN_DIR/flush-health" + if [ -s "$RUN_DIR/collector-progress" ]; then + read -r progress_owner extra extra received errors flush_errors <"$RUN_DIR/collector-progress" + fi + now="$(log_clock)"; state=stopped + if log_process_info "$owner" >/dev/null; then + state=starting + if [ -n "$birth" ] && [ "$(log_process_info "$reader")" = "$birth" ]; then + state=waiting-for-events + if [ -s "$RUN_DIR/collector-progress" ]; then + read -r progress_owner extra extra received errors flush_errors <"$RUN_DIR/collector-progress" + if [ "$progress_owner" = "$owner" ] && [ "$received" -gt 0 ]; then + state=idle-unverified + [ "$input_up" -gt 0 ] && [ "$now" -ge "$input_up" ] && [ "$((now - input_up))" -le 120 ] && state=receiving + fi + fi + else state=reader-stopped; fi + fi + printf 'collector_state=%s\ncollector_pid=%s\nreader_pid=%s\nreceived_lines=%s\nlast_input_epoch=%s\nlast_event_epoch=%s\naccepted_events=%s\nsuppressed_events=%s\nappend_errors=%s\nflush_errors=%s\nlast_flush_epoch=%s\nheartbeat_sample_seconds=%s\n' "$state" "$owner" "$reader" "$received" "$last_input" "$last_event" "$accepted" "$suppressed" "$errors" "$flush_errors" "$flush_epoch" "$LOG_HEARTBEAT_SECONDS" + if [ "$input_up" -gt 0 ] && [ "$now" -ge "$input_up" ]; then printf 'last_input_age_seconds=%s\n' "$((now - input_up))"; fi + if [ "$event_up" -gt 0 ] && [ "$now" -ge "$event_up" ]; then printf 'last_event_age_seconds=%s\n' "$((now - event_up))"; fi + printf 'rejected_events=%s\nlast_commit_epoch=%s\n' "$failed" "$(cat "$RUN_DIR/commit-health" 2>/dev/null || printf 0)" + printf 'runtime_bytes=%s\n' "$(du -sk "$RUN_DIR" 2>/dev/null | awk '{print $1 * 1024}')" + for extra in "$owner" "$reader"; do + case "$extra" in ''|*[!0-9]*) continue ;; esac + if [ -r "/proc/$extra/status" ]; then + awk -v pid="$extra" '/^VmRSS:/ {print "process_" pid "_rss_kib=" $2}' "/proc/$extra/status" + fi + done +} +log_collect_line() { + # Explicit stdin is essential: awk/redaction must never inherit the live FIFO + # and drain subsequent syslog records behind the collector's read loop. + printf '%s\n' "$1" | log_append +} log_collect() { - local owner + # Lifecycle variables belong to this collect process: the EXIT trap runs after + # this function returns, when function-local variables have gone out of scope. log_open || return 1 log_close - if ! mkdir "$RUN_DIR/collector-lock" 2>/dev/null; then - owner="$(cat "$RUN_DIR/collector-lock/pid" 2>/dev/null)" - case "$owner" in ''|*[!0-9]*) ;; *) kill -0 "$owner" 2>/dev/null && return 0 ;; esac - rm -f "$RUN_DIR/collector-lock/pid"; rmdir "$RUN_DIR/collector-lock" 2>/dev/null || return 1 - mkdir "$RUN_DIR/collector-lock" || return 1 - fi - log_self_pid >"$RUN_DIR/collector-lock/pid" - rm -f "$RUN_DIR"/stream.* - fifo="$RUN_DIR/stream.$$"; reader=''; writer=''; timer='' - trap 'trap - EXIT; for child in "$reader" "$writer" "$timer"; do [ -z "$child" ] || kill "$child" 2>/dev/null || true; done; wait; rm -f "$fifo" "$RUN_DIR/collector-lock/pid"; rmdir "$RUN_DIR/collector-lock" 2>/dev/null || true; (log_flush force) || true' EXIT + reader=''; birth='' + log_collector_lock; rc=$? + [ "$rc" != 2 ] || return 0 + [ "$rc" = 0 ] || return 1 + collector="$(cat "$RUN_DIR/collector-lock/pid")" + fifo="$RUN_DIR/stream.$collector" + received=0; append_errors=0; flush_errors=0; last_tick=0; fragment='' + trap 'log_collector_cleanup "$?"' EXIT trap 'exit 0' INT TERM - mkfifo "$fifo" || return 1 - logread -f -e 'owrt-remote\(\[[0-9][0-9]*\]\)\{0,1\}:' >"$fifo" 2>/dev/null & reader=$! - while IFS= read -r line; do printf '%s\n' "$line" | log_append; done <"$fifo" & writer=$! + # Only the owner may remove stale runtime FIFOs. Persistent chunks are untouched. + rm -f "$RUN_DIR"/stream.* + rmdir "$RUN_DIR/diag-lock" 2>/dev/null || true + mkfifo "$fifo" || { log_diag fifo-create-failed; return 1; } + # A single parent reads and appends. No background writer can silently die or + # consume signals while procd still observes a live supervising shell. + exec 3<>"$fifo" || { log_diag fifo-open-failed; return 1; } + logread -f -l 64 -e 'owrt-remote\(\[[0-9][0-9]*\]\)\{0,1\}:' 3>&- >"$fifo" 2>/dev/null & reader=$! + birth="$(log_process_info "$reader")" + printf '%s %s\n' "$reader" "$birth" >"$RUN_DIR/collector-lock/reader" + log_collector_progress || return 1 while :; do - # Lock-operation traps must not replace the collector's child/FIFO cleanup. - (log_flush) || true - sleep 5 & timer=$! - wait "$timer" || return 0 - kill -0 "$reader" 2>/dev/null && kill -0 "$writer" 2>/dev/null || return 1 + line=''; rc=0 + # BusyBox ash supports timed, bounded reads. The FIFO itself stays on tmpfs. + IFS= read -r -t 5 -n 4096 line <&3 || rc=$? + if [ "$rc" = 0 ] || [ -n "$line" ]; then + # A logd line is <= its fixed line buffer. Drop an oversized fragment + # instead of growing an unbounded shell string on malformed input. + if [ "$rc" != 0 ] || [ "${#line}" -ge 4096 ]; then + fragment=1; log_diag fragmented-input + elif [ -n "$fragment" ]; then + fragment='' + else + received=$((received + 1)) + log_collector_progress || return 1 + if ! (log_collect_line "$line"); then + append_errors=$((append_errors + 1)); log_collector_progress || true + log_diag collector-append-failed + # Do not pretend to be healthy after an unhandled write failure. + return 1 + fi + fi + fi + now="$(log_clock)" + if [ "$now" -lt "$last_tick" ] || [ "$((now - last_tick))" -ge 5 ]; then + last_tick="$now" + if ! (log_flush); then flush_errors=$((flush_errors + 1)); log_diag collector-flush-failed; fi + log_collector_progress || return 1 + fi + if ! log_process_info "$reader" >/dev/null; then log_diag reader-exited; return 1; fi + # A timeout with no input is normal idle, not a failure or proof of health. done } case "${1:-}" in check) log_open && log_close ;; status) log_status ;; + diagnostics) log_init && { [ ! -f "$RUN_DIR/diagnostics" ] || cat "$RUN_DIR/diagnostics"; } ;; migrate) log_migrate ;; flush) log_flush "${2:-}" ;; collect) log_collect ;; @@ -729,5 +944,5 @@ case "${1:-}" in event) case "${3:-ok}" in bad|error) level=ERROR ;; warn) level=WARN ;; *) level=OK ;; esac { printf '%s owrt-remote: %s\n' "$(date '+%Y-%m-%d %H:%M:%S')" "$2"; cat; } | log_append "$level" ;; - *) printf 'Usage: logs.sh collect|read [live]|clear|append|event MESSAGE|status|flush [force]\n' >&2; exit 2 ;; + *) printf 'Usage: logs.sh collect|read [live]|clear|append|event MESSAGE|status|diagnostics|flush [force]\n' >&2; exit 2 ;; esac diff --git a/tests/LOG_COLLECTOR.md b/tests/LOG_COLLECTOR.md new file mode 100644 index 0000000..d0fe022 --- /dev/null +++ b/tests/LOG_COLLECTOR.md @@ -0,0 +1,111 @@ +# Journal collector liveness fix (local v117) + +## What the reported evidence does and does not prove + +The previous `log_new_fact_locked()` normalized syslog timestamps and stored the +first successful `Heartbeat: OK; ...` in `heartbeat-ok`. The identical success +was then suppressed **without an expiry**, even after hours. A 136-byte chunk, +`ack=total`, empty pending and live logread therefore do not prove a stalled +collector. They also do not establish a memory leak. A shell syntax check and +`persistent=1` prove neither event delivery nor process health. + +The old supervisor checked only `kill -0` for its background writer and reader; +it ignored `log_append` failure. This could conceal write/lock errors while the +supervisor remained alive. These failure paths are corrected independently of +the deduplication issue. No claim is made that they caused this particular +AX3000T report without measurements from that device. + +## Changes + +- One supervising collect shell consumes its private FIFO with BusyBox ash + timed reads. There is one logread child, no separate persistent writer shell + and no five-second sleep child. At most one 4096-byte input fragment is kept + in a shell variable. Oversized/fragmented input is diagnosed and discarded; + it cannot grow a shell string without a bound. +- Each complete line is handed to `log_append` over an explicit short pipe, + isolated from the live FIFO. Append failure is checked, recorded and causes + a nonzero exit so procd can recover. Lock-operation traps run in subshells + and cannot replace the collector's cleanup trap. +- Normal stop terminates/reaps logread, closes/removes the FIFO, flushes a final + bounded batch and releases only its own collector lock. SIGKILL recovery + recognizes a stale owner and terminates its recorded orphan reader only if + PID **and kernel start time** match. Stale-lock reaping is serialized. +- procd's journal instance retries every 10 seconds without exhausting a + five-crash budget. The archive instance and networking services are unchanged. +- Builds without fractional BusyBox sleep use whole-second waits. Lock and + child cleanup deadlines account for that duration; unsupported `sleep 0.1` + cannot produce a fast loop, repeated errors or a 100-second lock wait. +- A successful heartbeat is saved at most once per 300 seconds when unchanged, + plus changes and recovery. Intervening duplicates update small RAM health + metadata only. Flash flush is still once per 60 seconds, <=16 KiB per batch, + with existing size/age/part limits. No per-heartbeat Flash write is introduced. +- Starting logread requests only a 64-message system-ring tail plus follow, + to recover a short interruption. Existing fact dedup and bounded pending + apply; this is not a guarantee of lossless collection during a long outage. +- `status` adds input and accepted-event times, counters, collector state, + append/flush errors, the last completed batch/commit, allocated runtime bytes + and RSS for the collector/reader when proc provides it. `receiving` means + recently processed input; `idle-unverified` explicitly does not assert health. +- `diagnostics` is a private RAM-only <=4 KiB reason-code history (plus <=2 KiB + compaction), with duplicate-rate limiting. It records FIFO creation/open, + fragmented input, lock timeout, append, flush and reader exit failures without + storing event text, keys or command lines. It is not sent back through logger, + avoiding diagnostic feedback loops. +- Existing validated persistent storage, immutable checksummed chunks, ack + protocol and limited RAM fallback remain. Runtime health is lost on reboot; + retained journal/ack data are not cleared by a collector restart. + +## Local install and rollback (Windows PowerShell 5.1/11) + +The dedicated installer replaces **only** `logs.sh` and `owrt-remote-logs`. +It backs up code/service state, preserves journal chunks and UCI configuration, +restarts only the log service, verifies a real logger marker through the FIFO +and performs one explicit deployment flush. Failure restores the previous code. +It does not remove `xray.zip`, update Xray, restart the tunnel, modify Podkop, +WAN/DNS or policy routing. VPS files and hosting URLs are untouched. + +```powershell +powershell.exe -NoProfile -ExecutionPolicy Bypass -File "C:\Users\k.zolotarev95\Documents\openwrt overlay\vps\deploy-log-collector-fix.ps1" -Router "root@192.168.2.1" -Action Install +``` + +Retained backup: `/root/owrt-remote-backup-before-log-collector-v117`. +Rollback leaves journal data intact: + +```powershell +powershell.exe -NoProfile -ExecutionPolicy Bypass -File "C:\Users\k.zolotarev95\Documents\openwrt overlay\vps\deploy-log-collector-fix.ps1" -Router "root@192.168.2.1" -Action Rollback +``` + +After installation compare two samples a minute apart. `last_input_epoch` and +`suppressed_events` should advance on stable heartbeat. An unchanged heartbeat +becomes a new retained sample after five minutes. A distinct marker is visible +immediately in read, then committed by the next normal flush (up to ~65 seconds). + +```powershell +ssh root@192.168.2.1 'logger -t owrt-remote collector-user-check; sleep 3; sh /usr/lib/owrt-remote/logs.sh read | tail -n 12; sh /usr/lib/owrt-remote/logs.sh status; sh /usr/lib/owrt-remote/logs.sh diagnostics; free' +``` + +Measure memory using `MemAvailable`, tmpfs allocation and process RSS over time; +`free` memory alone includes reclaimable kernel/filesystem cache. Do not attribute +all process RSS or low available RAM to the log spool without these measurements. +No router or VPS is modified by automated tests. + +## Automated coverage and limits + +- Real FIFO tests using MSYS shell on Windows: live heartbeat sampling, distinct + events through pending/Flash, duplicate collector rejection, reader death, + append failure, owner SIGKILL and successor recovery, TERM cleanup/final flush. +- Real BusyBox ash/applets: spool bounds, atomic/checksum/ack recovery, sampling, + limited diagnostics, fallback and cleanup paths; Windows BusyBox lacks FIFOs. +- Sandboxed installer: install, first-backup preservation, rollback, bad syntax, + failed verification restoration, journal retention. +- Existing redaction/CGI/Hub transport regressions and shell syntax checks. + +This is not a claim of physical AX3000T / OpenWrt 24.10.3 UBIFS testing, hardware +power-cut testing or a measured RAM reduction in MiB. The actual device still +needs event/flush/reboot and memory observations before release. + +Implementation references checked: +[OpenWrt 24.10 ubox version](https://github.com/openwrt/openwrt/blob/openwrt-24.10/package/system/ubox/Makefile), +[its logread](https://github.com/openwrt/ubox/blob/85f1053019caf4cd333795760950235ee4529ba7/log/logread.c), +[BusyBox ash timed read](https://github.com/mirror/busybox/blob/1_36_stable/shell/ash.c), +[procd respawn](https://github.com/openwrt/procd/blob/master/service/instance.c). diff --git a/tests/test_global_logs.py b/tests/test_global_logs.py index 2eec413..5044fe1 100644 --- a/tests/test_global_logs.py +++ b/tests/test_global_logs.py @@ -108,7 +108,7 @@ class GlobalLogTests(unittest.TestCase): self.helper_call('append', good * 3) self.helper_call('append', 'Other action\n') self.helper_call('append', good, env={**self.env, 'TEST_UPTIME': '2000'}) - self.assertEqual(self.helper_call('read', env={**self.env, 'TEST_UPTIME': '2000'}).count('Heartbeat: OK;'), 1) + self.assertEqual(self.helper_call('read', env={**self.env, 'TEST_UPTIME': '2000'}).count('Heartbeat: OK;'), 2) failed = 'Heartbeat: ошибка; endpoint=https://hub.example; rc=7; connection refused\n' self.helper_call('append', failed * 3, env={**self.env, 'TEST_UPTIME': '2001'}) self.helper_call('append', good, env={**self.env, 'TEST_UPTIME': '2002'}) @@ -117,7 +117,7 @@ class GlobalLogTests(unittest.TestCase): env={**self.env, 'TEST_UPTIME': '2004'}) stored = self.helper_call('read', env={**self.env, 'TEST_UPTIME': '2004'}) self.assertEqual(stored.count('[ERROR] Heartbeat:'), 2, 'same failure after recovery must appear') - self.assertEqual(stored.count('[OK] Heartbeat:'), 3, 'first, recovery and another WAN') + self.assertEqual(stored.count('[OK] Heartbeat:'), 4, 'first, periodic sample, recovery and another WAN') def test_recurring_failure_is_recorded_again_after_a_quiet_period(self): failure = 'owrt-remote: cannot download Xray\n' diff --git a/tests/test_log_collector.py b/tests/test_log_collector.py new file mode 100644 index 0000000..16045c2 --- /dev/null +++ b/tests/test_log_collector.py @@ -0,0 +1,172 @@ +"""Real FIFO lifecycle tests; MSYS provides FIFOs on Windows, BusyBox tests the applets separately.""" +import os +from pathlib import Path +import subprocess +import time +import unittest + +import test_persistent_logs as spool + + +class CollectorTests(unittest.TestCase): + setUp = spool.PersistentLogsTests.setUp + call = spool.PersistentLogsTests.call + status = spool.PersistentLogsTests.status + flush = spool.PersistentLogsTests.flush + journal = spool.PersistentLogsTests.journal + # Reuse filesystem fixtures without inheriting the persistent spool test cases. + def start(self, extra=""): + self.input = self.root / 'input' + self.env['TEST_INPUT'] = self.input.as_posix() + self.env['TEST_CLOCK_FILE'] = (self.root / 'clock').as_posix() + (self.root / 'clock').write_text('1000\n') + mocks = r''' +log_clock() { cat "$TEST_CLOCK_FILE"; } +logread() { + exec 4<>"$TEST_INPUT" + while :; do if IFS= read -r -t 1 item <&4; then printf '%s\n' "$item"; fi; done +} +''' + source = self.source.replace('case "${1:-}" in', mocks + extra + '\ncase "${1:-}" in', 1) + self.helper.write_text(source, encoding='utf-8', newline='\n') + if not getattr(self, 'input_created', False): + result = subprocess.run([*self.shell, '-c', 'mkfifo "$TEST_INPUT"'], env=self.env, + capture_output=True, timeout=10) + if result.returncode: + self.skipTest('This shell has no FIFO support') + self.input_created = True + output = (self.root / ('process-%d.log' % len(getattr(self, 'processes', [])))).open('wb') + proc = subprocess.Popen([*self.shell, self.helper.as_posix(), 'collect'], env=self.env, + stdout=output, stderr=output) + if not hasattr(self, 'processes'): + self.processes = [] + self.addCleanup(self.stop_processes) + self.processes.append((proc, output)) + return proc + + def wait_for(self, predicate, seconds=20): + deadline = time.monotonic() + seconds + while time.monotonic() < deadline: + value = predicate() + if value: + return value + time.sleep(.15) + self.fail('Collector did not reach expected state') + + def ready(self): + return self.wait_for(lambda: (self.run_dir / 'collector-lock/reader').exists()) + + def signal(self, pid, signal='TERM'): + return subprocess.run([*self.shell, '-c', 'kill -"$1" "$2"', '--', signal, str(pid)], + env=self.env, capture_output=True, timeout=10) + + def stop_processes(self): + lock = self.run_dir / 'collector-lock/pid' + if lock.exists(): + self.signal(lock.read_text().strip()) + for proc, output in self.processes: + try: + proc.wait(timeout=15) + except subprocess.TimeoutExpired: + proc.kill() + proc.wait(timeout=5) + output.close() + + def send(self, *lines): + result = subprocess.run([*self.shell, '-c', 'cat >"$TEST_INPUT"'], input='\n'.join(lines) + '\n', + encoding='utf-8', env=self.env, capture_output=True, timeout=15) + self.assertEqual(result.returncode, 0, result.stderr) + + def test_actual_fifo_heartbeat_freshness_sample_and_batch_flush(self): + proc = self.start(); self.ready() + good = 'Thu Oct 15 22:52:36 2026 user.notice owrt-remote: Heartbeat: OK; Hub=https://hub.example; WAN_direct=1; WAN_active=wan' + self.send(good) + self.wait_for(lambda: self.status()['accepted_events'] == '1') + self.send(good.replace('22:52:36', '22:53:08')) + self.wait_for(lambda: self.status()['suppressed_events'] == '1') + state = self.status() + self.assertEqual(state['collector_state'], 'receiving') + self.assertEqual(state['received_lines'], '2') + self.assertGreater(int(state['last_input_epoch']), 0) + self.assertEqual(state['committed_bytes'], '0') + (self.root / 'clock').write_text('1301\n') + self.send(good.replace('22:52:36', '22:57:38'), 'owrt-remote: panel action after heartbeat') + self.wait_for(lambda: self.status()['accepted_events'] == '3') + read = self.call('read').stdout + self.assertEqual(read.count('Heartbeat: OK;'), 2) + self.assertIn('panel action after heartbeat', read) + self.flush() + self.assertEqual(self.status()['pending_bytes'], '0') + self.assertIn(b'panel action after heartbeat', b''.join(p.read_bytes() for p in self.journal.glob('part.*'))) + self.assertIsNone(proc.poll()) + + def test_duplicate_collect_does_not_remove_owner_fifo(self): + proc = self.start(); self.ready() + owner = (self.run_dir / 'collector-lock/pid').read_text() + fifo = list(self.run_dir.glob('stream.*')) + other = self.start(); self.assertEqual(other.wait(timeout=15), 0) + self.assertEqual((self.run_dir / 'collector-lock/pid').read_text(), owner) + self.assertEqual(list(self.run_dir.glob('stream.*')), fifo) + self.send('owrt-remote: owner still receives') + self.wait_for(lambda: self.status()['accepted_events'] == '1') + self.assertIsNone(proc.poll()) + + def test_reader_death_exits_nonzero_and_retains_journal_for_restart(self): + proc = self.start(); self.ready() + self.send('owrt-remote: before logread failure') + self.wait_for(lambda: self.status()['accepted_events'] == '1') + reader = (self.run_dir / 'collector-lock/reader').read_text().split()[0] + self.signal(reader, 'KILL') + self.assertNotEqual(proc.wait(timeout=20), 0) + self.assertFalse((self.run_dir / 'collector-lock').exists()) + self.assertFalse(list(self.run_dir.glob('stream.*'))) + self.assertIn('reader-exited', self.call('diagnostics').stdout) + self.assertIn('before logread failure', self.call('read').stdout) + new = self.start(); self.ready() + self.send('owrt-remote: after reader recovery') + self.wait_for(lambda: self.status()['accepted_events'] == '2') + self.assertIsNone(new.poll()) + + def test_sigkill_owner_reaps_identified_reader_and_stale_fifo(self): + proc = self.start(); self.ready() + owner = (self.run_dir / 'collector-lock/pid').read_text().strip() + self.send('owrt-remote: retained before killed collector') + self.wait_for(lambda: self.status()['accepted_events'] == '1') + old_reader = (self.run_dir / 'collector-lock/reader').read_text().split()[0] + self.signal(owner, 'KILL'); proc.wait(timeout=10) + new = self.start() + self.wait_for(lambda: (self.run_dir / 'collector-lock/pid').exists() and + (self.run_dir / 'collector-lock/pid').read_text().strip() != owner) + self.wait_for(lambda: (self.run_dir / 'collector-lock/reader').exists() and + (self.run_dir / 'collector-lock/reader').read_text().split()[0] != old_reader) + self.assertEqual(len(list(self.run_dir.glob('stream.*'))), 1) + self.send('owrt-remote: recovered collector') + self.wait_for(lambda: self.status()['accepted_events'] == '2') + text = self.call('read').stdout + self.assertIn('retained before killed collector', text) + self.assertIn('recovered collector', text) + self.assertIsNone(new.poll()) + + def test_append_failure_is_not_silently_ignored(self): + proc = self.start('\nlog_collect_line() { return 1; }\n'); self.ready() + self.send('owrt-remote: rejected input') + self.assertNotEqual(proc.wait(timeout=20), 0) + self.assertIn('collector-append-failed', self.call('diagnostics').stdout) + self.assertEqual(self.status()['append_errors'], '1') + self.assertFalse((self.run_dir / 'collector-lock').exists()) + + def test_term_flushes_and_cleans_children_fifo_and_owner_lock(self): + proc = self.start(); self.ready() + self.send('owrt-remote: shutdown flush') + self.wait_for(lambda: self.status()['accepted_events'] == '1') + owner = (self.run_dir / 'collector-lock/pid').read_text().strip() + self.signal(owner) + self.assertEqual(proc.wait(timeout=15), 0) + self.assertEqual(self.status()['pending_bytes'], '0') + self.assertIn('shutdown flush', self.call('read').stdout) + self.assertFalse(list(self.run_dir.glob('stream.*'))) + self.assertFalse((self.run_dir / 'collector-lock').exists()) + + +if __name__ == '__main__': + unittest.main() diff --git a/tests/test_log_collector_deploy.py b/tests/test_log_collector_deploy.py new file mode 100644 index 0000000..e40ed33 --- /dev/null +++ b/tests/test_log_collector_deploy.py @@ -0,0 +1,95 @@ +"""Deploy only the journal helper/service, preserving data and networking.""" +import os +from pathlib import Path +import re +import shlex +import shutil +import subprocess +import tempfile +import unittest + +ROOT = Path(__file__).resolve().parents[1] +GIT_BIN = Path(os.environ.get('ProgramFiles', 'C:/Program Files')) / 'Git/usr/bin' +SH = str(GIT_BIN / 'sh.exe') if os.name == 'nt' else shutil.which('sh') + +class CollectorDeploymentTests(unittest.TestCase): + def setUp(self): + self.tmp = tempfile.TemporaryDirectory(prefix='collector-deploy-') + self.addCleanup(self.tmp.cleanup) + self.root = Path(self.tmp.name) + self.targets = {'logs.sh': '/usr/lib/owrt-remote/logs.sh', 'logs-init': '/etc/init.d/owrt-remote-logs'} + self.initial = {} + for item, target in self.targets.items(): + path = self.root / target.lstrip('/') + path.parent.mkdir(parents=True, exist_ok=True) + path.write_text('#!/bin/sh\necho old-' + item + '-$1\n', newline='\n') + self.initial[item] = path.read_bytes() + self.backup = self.root / 'backup' + self.journal = self.root / 'retained-journal'; self.journal.write_text('keep history') + text = (ROOT / 'vps/deploy-log-collector-fix.ps1').read_text(encoding='utf-8') + self.script = text.split("$RemoteScript = @'\n", 1)[1].split("\n'@", 1)[0] + mapping = {target: (self.root / target.lstrip('/')).as_posix() for target in self.targets.values()} + mapping['/root/owrt-remote-backup-before-log-collector-v117'] = self.backup.as_posix() + pattern = '|'.join(re.escape(path) for path in sorted(mapping, key=len, reverse=True)) + self.script = re.sub(pattern, lambda m: mapping[m.group()], self.script) + mocks = r''' +ubus() { printf '{"running":true}\n'; } +jsonfilter() { printf true; } +logread() { return 0; } +mkfifo() { return 0; } +logger() { printf '%s\n' "$3" >"$TEST_MARKER"; } +sleep() { return 0; } +''' + self.script = self.script.replace('trap cleanup EXIT', mocks + '\ntrap cleanup EXIT') + self.env = {**os.environ, 'PATH': str(GIT_BIN) + os.pathsep + os.environ['PATH'], + 'TEST_MARKER': (self.root / 'marker').as_posix()} + + def deploy(self, action, bad=False, fail=False): + stage = self.root / 'stage'; stage.mkdir() + if action == 'Install': + helper = r'''#!/bin/sh +# log_collector_progress() +# LOG_HEARTBEAT_SECONDS=300 +case "$1" in +read) cat "$TEST_MARKER" ;; +status) echo persistent=1 ;; +esac +''' + if bad: helper = '#!/bin/sh\nif true; then\n' + (stage / 'logs.sh').write_text(helper, newline='\n') + (stage / 'logs-init').write_text('#!/bin/sh\necho new-service-$1\n', newline='\n') + script = self.script.replace('stage=/tmp/__UPLOAD_NAME__', 'stage=' + shlex.quote(stage.as_posix())) + script = script.replace('__UPLOAD_NAME__', 'sandbox').replace('__ACTION__', action) + if fail: script = script.replace('cmp "$source_dir/$item" "$path"', 'false') + result = subprocess.run([SH], input=script, text=True, encoding='utf-8', capture_output=True, env=self.env, timeout=20) + self.assertFalse(stage.exists()) + self.assertEqual(self.journal.read_text(), 'keep history') + return result + + def test_install_and_rollback_keep_first_backup_and_journal(self): + for _ in range(2): + result = self.deploy('Install') + self.assertEqual(result.returncode, 0, result.stderr) + self.assertIn('LOG_COLLECTOR_ROUTER_Install_OK', result.stdout) + for item, initial in self.initial.items(): self.assertEqual((self.backup / item).read_bytes(), initial) + result = self.deploy('Rollback') + self.assertEqual(result.returncode, 0, result.stderr) + for item, target in self.targets.items(): self.assertEqual((self.root / target.lstrip('/')).read_bytes(), self.initial[item]) + + def test_missing_backup_is_noop(self): + result = self.deploy('Rollback') + self.assertNotEqual(result.returncode, 0) + self.assertIn('Nothing changed', result.stderr) + + def test_invalid_shell_rejected_before_backup(self): + result = self.deploy('Install', bad=True) + self.assertNotEqual(result.returncode, 0) + self.assertFalse(self.backup.exists()) + + def test_failed_verify_restores_previous_code(self): + result = self.deploy('Install', fail=True) + self.assertNotEqual(result.returncode, 0) + self.assertIn('Restoring journal code', result.stderr) + for item, target in self.targets.items(): self.assertEqual((self.root / target.lstrip('/')).read_bytes(), self.initial[item]) + +if __name__ == '__main__': unittest.main() diff --git a/tests/test_persistent_logs.py b/tests/test_persistent_logs.py index f2da3a5..ef44b82 100644 --- a/tests/test_persistent_logs.py +++ b/tests/test_persistent_logs.py @@ -128,7 +128,7 @@ class PersistentLogsTests(unittest.TestCase): (self.run_dir / 'facts').write_text(''.join('1000\ts256-%035d\n' % i for i in range(63))) self.append('distinct event 63\ndistinct event 64\n') self.assertEqual(len((self.run_dir / 'facts').read_text().splitlines()), 64) - self.assertLessEqual((self.run_dir / 'facts').stat().st_size, 4096) + self.assertLessEqual((self.run_dir / 'facts').stat().st_size if (self.run_dir / 'facts').exists() else 0, 4096) def test_empty_digest_files_falsely_acked_by_old_hub_are_recovered(self): self.append('retained restart event\n'); self.flush() @@ -242,7 +242,7 @@ class PersistentLogsTests(unittest.TestCase): self.flush() self.assertEqual({p.name: p.read_bytes() for p in parts}, original) self.assertLessEqual((self.run_dir / 'pending').stat().st_size, 16384) - self.assertLessEqual((self.run_dir / 'facts').stat().st_size, 4096) + self.assertLessEqual((self.run_dir / 'facts').stat().st_size if (self.run_dir / 'facts').exists() else 0, 4096) first = self.export() self.acknowledge(first) self.flush() @@ -353,6 +353,32 @@ class PersistentLogsTests(unittest.TestCase): self.assertIn('after killed reaper', self.call('read').stdout) self.assertFalse(lock.exists()) + def test_integer_only_sleep_keeps_lock_timeout_bounded_and_quiet(self): + self.call('check') + source = self.source.split('case "${1:-}" in', 1)[0] + commands = r''' +log_init +mkdir "$RUN_DIR/lock" +printf '12345\n' >"$RUN_DIR/lock/pid" +sleep() { + printf '%s\n' "$1" >>"$RUN_DIR/pauses" + if [ "$1" = 0.1 ]; then echo "sleep: invalid number '0.1'" >&2; return 1; fi + [ "$1" = 1 ] +} +kill() { return 0; } +if log_lock; then exit 17; fi +''' + script = self.root / 'integer-sleep.sh' + script.write_text(source + commands, encoding='utf-8', newline='\n') + result = subprocess.run([*self.shell, script.as_posix()], env=self.env, + capture_output=True, text=True, timeout=15) + self.assertEqual(result.returncode, 0, result.stderr) + self.assertEqual(result.stderr, '') + pauses = (self.run_dir / 'pauses').read_text().splitlines() + self.assertEqual(pauses.count('0.1'), 1, 'probe once per process') + self.assertEqual(pauses.count('1'), 10, 'ten-second deadline, no busy spin or 100-second wait') + self.assertIn('lock-timeout', self.call('diagnostics').stdout) + def test_collector_failure_reaps_children_and_removes_fifo(self): # Actual FIFO integration is tested on POSIX; Windows BusyBox has no # FIFO support. Exercise its traps with mocked process/FIFO operations. @@ -373,7 +399,48 @@ wait() { return 0; } self.assertFalse(list(self.run_dir.glob('stream.*'))) self.assertFalse((self.run_dir / 'collector-lock').exists()) lines = (self.run_dir / 'kills').read_text().splitlines() - self.assertGreaterEqual(len([line for line in lines if not line.startswith('-0')]), 2) + self.assertGreaterEqual(len([line for line in lines if not line.startswith('-0')]), 1) + + def test_heartbeat_sampling_and_freshness_are_independent_of_flash_writes(self): + good = 'Thu Oct 15 22:52:36 2026 user.notice owrt-remote: Heartbeat: OK; Hub=https://hub.example; WAN_direct=1; WAN_active=wan\n' + self.append(good) + self.append(good.replace('22:52:36', '22:53:08'), TEST_UPTIME='1032') + state = self.status() + self.assertEqual((state['accepted_events'], state['suppressed_events']), ('1', '1')) + self.assertEqual(state['committed_bytes'], '0') + self.assertGreater(int(state['last_input_epoch']), 0) + self.assertEqual(state['collector_state'], 'stopped', 'persistent storage does not imply a running collector') + self.append(good.replace('22:52:36', '22:57:38'), TEST_UPTIME='1302') + self.assertEqual(self.call('read').stdout.count('Heartbeat: OK;'), 2) + self.flush() + self.assertEqual(self.status()['pending_bytes'], '0') + self.assertGreater(int(self.status()['last_commit_epoch']), 0) + + def test_private_diagnostics_are_bounded_rate_limited_and_no_syslog_feedback(self): + self.call('check') + source = self.source.split('case "${1:-}" in', 1)[0] + commands = r''' +log_init +logger() { echo forbidden-syslog-feedback >&2; return 1; } +i=0 +while [ "$i" -lt 140 ]; do + if [ "$((i % 2))" = 0 ]; then log_diag flush-failed; else log_diag lock-timeout; fi + i=$((i + 1)) +done +log_diag lock-timeout +''' + script = self.root / 'diag-test.sh' + script.write_text(source + commands, encoding='utf-8', newline='\n') + result = subprocess.run([*self.shell, script.as_posix()], env=self.env, + capture_output=True, text=True, timeout=30) + self.assertEqual(result.returncode, 0, result.stderr) + self.assertEqual(result.stderr, '') + diagnostics = self.run_dir / 'diagnostics' + self.assertLessEqual(diagnostics.stat().st_size, 4096) + self.assertFalse((self.run_dir / 'diagnostics.new').exists()) + self.assertFalse((self.run_dir / 'diag-lock').exists()) + self.assertEqual(self.call('diagnostics').stdout, diagnostics.read_text()) + self.assertFalse((self.journal / 'diagnostics').exists()) def test_legacy_migration_preserves_identity_ack_and_only_undelivered_bytes(self): legacy = self.root / 'legacy' diff --git a/vps/deploy-log-collector-fix.ps1 b/vps/deploy-log-collector-fix.ps1 new file mode 100644 index 0000000..8081885 --- /dev/null +++ b/vps/deploy-log-collector-fix.ps1 @@ -0,0 +1,148 @@ +param( + [ValidateNotNullOrEmpty()][string]$Router = 'root@192.168.2.1', + [ValidateSet('Install', 'Rollback')][string]$Action = 'Install', + [switch]$PrepareOnly +) + +$ErrorActionPreference = 'Stop' +$RepoRoot = Split-Path -Parent $PSScriptRoot +$UploadName = 'owrt-log-collector-' + [guid]::NewGuid().ToString('N') +$StageRoot = Join-Path ([System.IO.Path]::GetTempPath()) $UploadName +$Utf8 = [System.Text.UTF8Encoding]::new($false) +$Payloads = @{ + 'logs.sh' = 'files\usr\lib\owrt-remote\logs.sh' + 'logs-init' = 'files\etc\init.d\owrt-remote-logs' +} +$RemoteScript = @' +#!/bin/sh +set -eu +stage=/tmp/__UPLOAD_NAME__ +backup_dir=/root/owrt-remote-backup-before-log-collector-v117 +changed=0 +phase=preflight +items='logs.sh logs-init' +item_path() { + case "$1" in + logs.sh) printf /usr/lib/owrt-remote/logs.sh ;; + logs-init) printf /etc/init.d/owrt-remote-logs ;; + esac +} +complete_backup() { + [ -f "$1/complete" ] && [ -s "$1/logs.sh" ] && [ -s "$1/logs-init" ] +} +collector_running() { + ubus call service list '{"name":"owrt-remote-logs"}' | jsonfilter -e '@["owrt-remote-logs"].instances.journal.running' | grep -qx true +} +save_files() { + local item path + mkdir -p "$1" + chmod 700 "$1" + for item in $items; do path=$(item_path "$item"); cp -p "$path" "$1/$item" || return 1; done + if /etc/init.d/owrt-remote-logs enabled; then : >"$1/enabled"; fi + if collector_running; then : >"$1/running"; fi + : >"$1/complete" +} +restore_files() { + local item path + complete_backup "$1" || return 1 + /etc/init.d/owrt-remote-logs stop || true + for item in $items; do + path=$(item_path "$item") + cp -p "$1/$item" "$path.new-__UPLOAD_NAME__" && mv -f "$path.new-__UPLOAD_NAME__" "$path" || return 1 + done + if [ -f "$1/enabled" ]; then /etc/init.d/owrt-remote-logs enable; else /etc/init.d/owrt-remote-logs disable; fi + if [ -f "$1/running" ]; then /etc/init.d/owrt-remote-logs start; fi +} +cleanup() { + code=$? + trap - EXIT + if [ "$code" -ne 0 ]; then + printf 'Failed during %s (rc=%s).\n' "$phase" "$code" >&2 + if [ "$changed" = 1 ]; then + echo 'Restoring journal code/service state from this attempt; journal data retained.' >&2 + restore_files "$stage/previous" || echo 'Restore failed; retained backup is available.' >&2 + fi + fi + for item in $items; do path=$(item_path "$item"); rm -f "$path.new-__UPLOAD_NAME__"; done + rm -rf "$stage" + exit "$code" +} +trap cleanup EXIT +for item in $items; do path=$(item_path "$item"); test -s "$path"; done +command -v ubus >/dev/null +command -v jsonfilter >/dev/null +if [ '__ACTION__' = Install ]; then + for item in $items; do test -s "$stage/$item"; sh -n "$stage/$item"; done + command -v logread >/dev/null + command -v logger >/dev/null + command -v mkfifo >/dev/null + grep -Fq 'log_collector_progress()' "$stage/logs.sh" + grep -Fq 'LOG_HEARTBEAT_SECONDS=300' "$stage/logs.sh" + if [ ! -e "$backup_dir" ]; then save_files "$stage/retained"; mv "$stage/retained" "$backup_dir"; fi + complete_backup "$backup_dir" || { echo "Incomplete backup: $backup_dir. Nothing changed." >&2; exit 1; } + source_dir=$stage +else + complete_backup "$backup_dir" || { echo "No complete backup: $backup_dir. Nothing changed." >&2; exit 1; } + for item in $items; do sh -n "$backup_dir/$item"; done + source_dir=$backup_dir +fi +save_files "$stage/previous" +phase=install +changed=1 +/etc/init.d/owrt-remote-logs stop +for item in $items; do + path=$(item_path "$item") + cp "$source_dir/$item" "$path.new-__UPLOAD_NAME__" + case "$item" in logs.sh) chmod 644 "$path.new-__UPLOAD_NAME__" ;; *) chmod 755 "$path.new-__UPLOAD_NAME__" ;; esac + mv -f "$path.new-__UPLOAD_NAME__" "$path" +done +if [ '__ACTION__' = Install ]; then + /etc/init.d/owrt-remote-logs enable + /etc/init.d/owrt-remote-logs start + phase=verify-delivery + marker=collector-check-__UPLOAD_NAME__ + ready=0 + for attempt in 1 2 3 4 5 6 7 8 9 10; do + logger -t owrt-remote "$marker" + sleep 1 + if collector_running && sh /usr/lib/owrt-remote/logs.sh read | grep -Fq "$marker"; then ready=1; break; fi + done + [ "$ready" = 1 ] || { echo 'Collector did not capture the test event; restoring code.' >&2; exit 1; } + # A forced flush verifies the write path once at deployment, not each Heartbeat. + phase=verify-flush + sh /usr/lib/owrt-remote/logs.sh flush force + sh /usr/lib/owrt-remote/logs.sh read | grep -Fq "$marker" + sh /usr/lib/owrt-remote/logs.sh status + sh /usr/lib/owrt-remote/logs.sh diagnostics +else + if [ -f "$backup_dir/enabled" ]; then /etc/init.d/owrt-remote-logs enable; else /etc/init.d/owrt-remote-logs disable; fi + if [ -f "$backup_dir/running" ]; then /etc/init.d/owrt-remote-logs start; fi +fi +for item in $items; do path=$(item_path "$item"); cmp "$source_dir/$item" "$path"; done +printf 'Backup: %s\n' "$backup_dir" +printf '\nLOG_COLLECTOR_ROUTER_%s_OK\n' '__ACTION__' +'@ +$RemoteScript = $RemoteScript.Replace('__UPLOAD_NAME__', $UploadName).Replace('__ACTION__', $Action) +try { + [System.IO.Directory]::CreateDirectory($StageRoot) | Out-Null + [System.IO.File]::WriteAllText((Join-Path $StageRoot 'run.sh'), $RemoteScript.Replace("`r`n", "`n") + "`n", $Utf8) + if ($Action -eq 'Install') { + foreach ($Item in $Payloads.GetEnumerator()) { + $Content = [System.IO.File]::ReadAllText((Join-Path $RepoRoot $Item.Value)).Replace("`r`n", "`n") + [System.IO.File]::WriteAllText((Join-Path $StageRoot $Item.Key), $Content, $Utf8) + } + } + if ($PrepareOnly) { Write-Host "Prepared only: $StageRoot"; return } + & scp -O -r -- $StageRoot "${Router}:/tmp/" + if ($LASTEXITCODE -ne 0) { throw 'Upload failed. The installation has not started.' } + & ssh -T -o ConnectTimeout=10 $Router "sh /tmp/$UploadName/run.sh" + if ($LASTEXITCODE -ne 0) { throw "$Action failed. See the output above; success was not confirmed." } + Write-Host "Log collector fix $Action finished successfully." +} finally { + if (-not $PrepareOnly) { + $Resolved = [System.IO.Path]::GetFullPath($StageRoot) + $Expected = [System.IO.Path]::GetFullPath((Join-Path ([System.IO.Path]::GetTempPath()) $UploadName)) + if ($Resolved -ne $Expected -or (Split-Path -Leaf $Resolved) -ne $UploadName) { throw 'Unsafe staging cleanup path.' } + if (Test-Path -LiteralPath $Resolved) { Remove-Item -LiteralPath $Resolved -Recurse -Force } + } +}