@glassmkr/crucible 0.10.3 → 0.11.0

package/dist/lib/types.d.ts

@@ -31,12 +31,200 @@ export interface Snapshot {
     file_descriptors?: FileDescriptorData;
     expected_reboot?: boolean;
     expected_reboot_reason?: string;
+    /** EDAC memory-error counters per memory controller + DIMM. */
+    ecc_edac?: EdacSnapshot;
+    /** PSI pressure-stall counters per resource (cpu, memory, io). */
+    psi?: PsiSnapshot;
+    /** /proc/vmstat swap-in/out rates. */
+    vmstat?: VmstatSnapshot;
+    /** pstore / kdump / wtmp signals corroborating a reboot. */
+    reboot_evidence?: RebootEvidence;
+    /** Hardware RAID controllers scraped via vendor CLIs. */
+    hardware_raid?: HardwareRaidSnapshot;
+    /** Per-process FD scan (top-50 consumers + RLIMIT_NOFILE). */
+    process_fd?: ProcessFdSnapshot;
+    /** LACP / bonding driver state from /proc/net/bonding. */
+    bonding?: BondingSnapshot;
+    /** TCP segment / retransmit / listen-queue counters from
+     *  /proc/net/snmp + /proc/net/netstat. */
+    tcp_stats?: TcpStatsSnapshot;
+}
+export interface EdacDimm {
+    /** dimm_label (vendor-defined string, e.g. "CPU1_DIMM_A1"). */
+    label: string;
+    /** dimm_location (slot number / chip-channel ordering). */
+    location: string;
+    /** DIMM size in MB; null if /sys did not report. */
+    size_mb: number | null;
+    ce_count: number;
+    ue_count: number;
+}
+export interface EdacSnapshot {
+    /** Sum of ce_count across all memory controllers. */
+    edac_corrected_total: number;
+    /** Sum of ue_count across all memory controllers. */
+    edac_uncorrected_total: number;
+    /** Per-DIMM detail. Empty array on hosts where dimm metadata
+     *  isn't exposed (older EDAC drivers). */
+    dimms: EdacDimm[];
+}
+export interface PsiResource {
+    /** Rolling average % over the last 10 / 60 / 300 seconds. */
+    avg10: number;
+    avg60: number;
+    avg300: number;
+    /** Cumulative microseconds stalled since boot. */
+    total: number;
+}
+export interface PsiSnapshot {
+    cpu?: {
+        some: PsiResource;
+        full?: PsiResource;
+    };
+    memory?: {
+        some: PsiResource;
+        full?: PsiResource;
+    };
+    io?: {
+        some: PsiResource;
+        full?: PsiResource;
+    };
+}
+export interface VmstatSnapshot {
+    /** Cumulative pswpin since boot. */
+    pswpin_total: number;
+    pswpout_total: number;
+    /** Per-second swap-in rate over the most recent interval; null on
+     *  the first snapshot (no baseline) or after a counter reset (host
+     *  reboot mid-session). */
+    pswpin_rate: number | null;
+    pswpout_rate: number | null;
+}
+export interface RebootEvidence {
+    /** True if /sys/fs/pstore/ contains any dmesg-* / console-* records
+     *  from the prior kernel. */
+    pstore_present: boolean;
+    /** Number of pstore records found (zero when pstore_present=false). */
+    pstore_record_count: number;
+    /** True if /var/crash/ contains a kdump vmcore. */
+    vmcore_present: boolean;
+    /** Most recent `last reboot -F` output line, verbatim. Null if
+     *  `last` is unavailable or wtmp is empty. */
+    wtmp_reboot_record: string | null;
+    /** Heuristic: true when wtmp shows a `shutdown` record before the
+     *  most recent reboot (suggests a clean shutdown). false when only
+     *  the boot record is present (suggests hard reset or power loss). */
+    prior_shutdown_clean: boolean;
+}
+export interface HardwareRaidController {
+    vendor: "dell" | "hpe" | "lsi" | "adaptec";
+    controller_id: string;
+    /** Vendor-reported overall state, e.g. "Optimal", "Degraded",
+     *  "Critical", "Failed", or "Unknown". The dashboard's
+     *  raid_degraded evaluator pages on any state != "Optimal". */
+    state: string;
+    /** Count of physical disks the controller flagged as failed /
+     *  degraded; null when the parser couldn't extract this. */
+    degraded_disks: number | null;
+    /** Optional vendor-text excerpt the dashboard can surface in
+     *  evidence; null when not captured. */
+    raw_summary: string | null;
+}
+export interface HardwareRaidSnapshot {
+    controllers: HardwareRaidController[];
 }
 export interface ConntrackData {
     available: boolean;
     count: number;
     max: number;
     percent: number;
+    /** C9 (2026-05-19): cumulative insert_failed counter (sum across CPUs)
+     *  from /proc/net/stat/nf_conntrack. Optional because pre-0.11.0
+     *  agents omit it. */
+    insert_failed_total?: number;
+    /** C9: cumulative drop counter from /proc/net/stat/nf_conntrack. */
+    drop_total?: number;
+    /** Per-second insert_failed rate over the most recent snapshot
+     *  interval. Null on first snapshot, on counter reset, or when the
+     *  stat file is unavailable. */
+    insert_failed_rate_per_sec?: number | null;
+    drop_rate_per_sec?: number | null;
+}
+export interface ProcessFdEntry {
+    pid: number;
+    comm: string;
+    fd_count: number;
+    rlimit_nofile_soft: number;
+    rlimit_nofile_hard: number;
+    /** fd_count / rlimit_nofile_soft * 100, rounded to one decimal. Zero
+     *  when soft limit is unlimited (no useful proximity signal). */
+    percent_of_soft_limit: number;
+}
+export interface ProcessFdSnapshot {
+    available: boolean;
+    reason?: string;
+    /** Top 50 processes by fd_count. */
+    top_consumers: ProcessFdEntry[];
+    /** Number of numeric /proc/<pid> entries we considered. */
+    total_processes_scanned: number;
+    /** Aggregate signal: max percent_of_soft_limit across top_consumers.
+     *  Null when top_consumers is empty. */
+    highest_percent_of_limit: number | null;
+}
+export interface BondSlave {
+    name: string;
+    mii_status: string;
+    link_failure_count: number;
+    permanent_hw_addr: string;
+    aggregator_id: number | null;
+    partner_churn_state: string | null;
+    partner_lacp_port_state: number | null;
+    /** Convenience flag derived from the LACP port-state bitfield's
+     *  synchronization bit (bit 3, 0x08). Null when the bond is not
+     *  LACP or partner state was not captured. */
+    partner_lacp_synchronized: boolean | null;
+}
+export interface BondAggregator {
+    id: number;
+    number_of_ports: number;
+    actor_key: number | null;
+    partner_key: number | null;
+    partner_mac_address: string | null;
+}
+export interface Bond {
+    name: string;
+    mode: string;
+    is_lacp: boolean;
+    lacp_rate: string | null;
+    slaves: BondSlave[];
+    /** Equal to slaves.length; surfaces the "configured" port count
+     *  alongside active_aggregator.number_of_ports so the dashboard can
+     *  compute a shortfall. */
+    configured_port_count: number;
+    active_aggregator: BondAggregator | null;
+}
+export interface BondingSnapshot {
+    available: boolean;
+    reason?: string;
+    bonds: Bond[];
+}
+export interface TcpStatsSnapshot {
+    available: boolean;
+    reason?: string;
+    out_segs_total?: number;
+    retrans_segs_total?: number;
+    in_segs_total?: number;
+    /** Retransmits divided by segments sent over the most recent
+     *  interval. Range 0.0 - 1.0. Null on first snapshot or counter
+     *  reset. Zero when no outbound traffic in the interval. */
+    retrans_ratio?: number | null;
+    retrans_rate_per_sec?: number | null;
+    /** Optional listen-queue counters from /proc/net/netstat TcpExt.
+     *  Absent when /proc/net/netstat is not readable. */
+    listen_overflows_total?: number;
+    listen_drops_total?: number;
+    listen_overflows_rate_per_sec?: number | null;
+    listen_drops_rate_per_sec?: number | null;
 }
 export interface SystemdData {
     failed_units: string[];
@@ -58,6 +246,21 @@ export interface FileDescriptorData {
     max: number;
     percent: number;
 }
+export interface ZfsVdev {
+    /** Vdev name, e.g. "raidz2-0", "mirror-0", or a raw device for
+     *  single-device top-level stripes. */
+    name: string;
+    /** Vdev state from `zpool status` (ONLINE, DEGRADED, FAULTED,
+     *  REMOVED, SUSPENDED, UNAVAIL). */
+    state: string;
+    /** Redundancy class. C6 addition (2026-05-19): scaled vdev severity
+     *  matrix on the dashboard side depends on this so a DEGRADED
+     *  raidz1 (zero remaining tolerance) pages differently from a
+     *  DEGRADED raidz2 (one disk-fault budget left). */
+    redundancy_class: "mirror" | "raidz1" | "raidz2" | "raidz3" | "draid" | "stripe";
+    /** Number of child devices under this vdev in a non-ONLINE state. */
+    degraded_disks_count: number;
+}
 export interface ZfsPool {
     name: string;
     state: string;
@@ -66,6 +269,14 @@ export interface ZfsPool {
     scrub_repaired?: string;
     last_scrub_date?: string;
     scrub_never_run?: boolean;
+    /** Top-level data vdevs. Always present from collector v0.10.4+.
+     *  Dashboard tolerates absent (older agents) via capability gates. */
+    vdevs: ZfsVdev[];
+    /** Separate log (SLOG / ZIL) vdevs. Empty array on pools without
+     *  a SLOG configured. */
+    slog_vdevs: ZfsVdev[];
+    /** Cache (L2ARC) vdevs. Empty array on pools without L2ARC. */
+    l2arc_vdevs: ZfsVdev[];
 }
 export interface ZfsData {
     pools: ZfsPool[];

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@glassmkr/crucible",
-  "version": "0.10.3",
+  "version": "0.11.0",
   "description": "Lightweight bare metal server monitoring. IPMI, SMART, OS, network. Opinionated alerts.",
   "type": "module",
   "main": "./dist/index.js",