A release shaped by the people who run it: a calmer dashboard, stub zones for internal domains, mutes for findings you've decided are harmless, a support package you can attach to an issue, and a watchdog that restarts a node that stops answering and leaves a record of why.
Highlights
-
A calmer dashboard. The first page is now a row of plain numbers, one smoothed chart (hover for the exact values), and top lists with one line per entry; tap a count for the full name and what it stands for. Kubernetes service names and a pod's search-list lookups read as one short row. The previous dashboard is still there at
#/?view=legacy. The top clients bar shows each device's blocked, cached, and forwarded split; the top lists and the active-clients number follow the time range you pick; and the Clients page's first tab is Last hour, a rolling hour instead of the clock hour. Most of these came from Greg Pakes' pull requests, rebuilt for the new page. -
Stub zones. Make an internal domain a stub zone (
url = "recursive://"withstub_zoneandstub_servers) and TelltaleDNS resolves names under it itself, starting at that zone's own servers and following their delegations, instead of asking a forwarder that can't. Your internal view wins over the public one. From John Depon's pull request. -
Mute a kind of finding for a device. A device that does something harmless every hour (two TVs looking up the reverse name of every address on the network) no longer raises the same anomaly every hour.
[[anomaly_mute]], or Mute for this device… on a finding, names the devices or groups, the kinds, and an optional end. A muted finding is still recorded and listed (turn on Show muted), but it isn't counted intelltale_anomalies_total, the badge, theanomalyalert, or health;telltale_anomalies_muted_totalcounts them. A mute reaches every node of a cluster, and agents have their own scope for it (config:write:anomalies). -
A support package.
telltale support-package, Settings → System → Support package, orPOST /api/v1/system/support-packagemakes one.tar.gzfor an issue: build and platform facts, your settings without secrets, health, upstream state, a metrics snapshot, and the node's recent log. By default every address and name is replaced by a pseudonym that is the same all through the file, no query names are included, and the log holds warnings and errors only;standardandnoneare opt-in. Nothing is uploaded: you read it first.--offlinemakes one for a node that won't start. The bug report form asks for it. For this the node now keeps a small log of its own (the last 2,000 lines in memory and two rotating files of about 1 MiB under<data_dir>/logs/;[support] persist_logs = falseturns the files off). -
A watchdog. A thread outside the runtime checks every 5 seconds that the runtime can run a task and that the node's own UDP listener answers. If neither has passed for 60 seconds, it writes a dump under
<data_dir>/watchdog/(which check failed, every thread and what it waits on, each thread's stack, the last log lines) and exits with status 70, so systemd, Compose, or Kubernetes restarts the node: a few seconds instead of an outage until someone notices. At most 6 restarts an hour.[watchdog]hasstall_secs,action = "dump"(write the dump and keep running),max_restarts_per_hour, andenabled. The newest dump goes into a support package. -
Debug files. Every release and edge build now publishes
telltale-<arch>-linux.debug.zstnext to each binary, under the signedSHA256SUMS. With one, a stack from a stripped binary (a watchdog dump, orgdb) becomes function names and source lines:zstd -d … -o telltale.debug, thenaddr2line -e telltale.debug -f -C -i <address>. -
Quick rules can be edited. Edit on the Quick rules page opens the rule filled in, shows the same preview as a new one, and saves it under the same ID. The rule keeps who made it and when, and keeping the current end time is the default.
-
Clusters.
- Remove a node that is gone (a machine you rebuilt comes back under a new ID): Remove on the Cluster page,
telltale ctl remove-node, orDELETE /api/v1/cluster/nodes/{id}. A node that is still up needs--force. - A Git-managed node takes over from an emergency primary. With automatic failover and the
gitopsauthority, when the Git-managed node returns it asks the emergency primary to hand over and is elected, with DNS answered throughout. - A replica that attached to another replica keeps looking for the primary and moves back to it.
- Upstream health in the web UI reads Healthy, Benched, or Probing (the API, metrics, and alerts keep the circuit-breaker values).
- Remove a node that is gone (a machine you rebuilt comes back under a new ID): Remove on the Cluster page,
-
Also new: an FAQ page on the project site; Dependabot, and CodeQL with test code left out of it.
Fixes
- Top clients names a device the way the query log does, including names from your router, mDNS, and other nodes.
- Query log, Live: a late page of results no longer fills the live view.
recursive://needs no bootstrap servers.- Helm: a warning when
alsoServemeetsexternalTrafficPolicy: Localwith resolver pods. - Dependencies: the 2026-11 RustSec advisories are cleared (among them rustls, zstd, and SQLite, now 3.54). Existing state databases open unchanged.
Upgrading from 0.6
Most installs need no change. Check these if they apply to you:
- Clusters: upgrade replicas first, then the primary, and make no
[[anomaly_mute]]until every node runs 0.7: an older node doesn't know the section, so it keeps its last configuration and logsupgrade this node. - The watchdog is on by default. A node that stops answering for a minute exits with status 70. systemd (
Restart=on-failure), Compose (restart: unless-stopped), and Kubernetes restart it; a node started by hand in a terminal just exits. Set[watchdog] action = "dump"to only write the dump, orenabled = falseto turn it off. Both are restart-only and node-local. - A small log on disk. The support package's log files under
<data_dir>/logs/are on by default (about 21 KB a day on a quiet node; a read-only data directory is fine, and[support] persist_logs = falseturns them off). - The dashboard has changed. Bookmark
#/?view=legacyif you want the old one.GET /api/v1/stats/toptakesfrom(andto), and the summary'sactiveClientsnow counts the range you ask for instead of this hour. The Clients page's Last hour tab keeps the address#/clients?tab=seen. - Quick rules: a write that replaces a rule (the same ID) keeps its
createdByandcreated; it no longer refreshes them. - Releases have three more files each (the debug files).
install.shandself-updatetake only the binary, as before.
Known issues
- A process hang (#7). Its cause isn't known yet. The watchdog restarts the node and the dump records what every thread was doing; with the debug file, that is readable. Stack capture is tested on x86_64 and arm64; on armv7 (32-bit) it builds and links but hasn't been tried.
- A replica attached to another replica still doesn't move back while the primary has a stream to it (#43).
- A group with one upstream SERVFAILs a name the upstream takes much longer than usual to answer (#35).
- The support package covers the node that answered. Asking every node (
--cluster) is planned (#55). Mutes match by device, address, network, or group, not by MAC address, and not by domain. - DNSSEC
validatemode: right after a start, an upstream that closes fresh DoT connections can make the first validations fail, so a signed name may get SERVFAIL until its keys are cached. Keeppermissiveorofffor now; see ADR-098. _dns.resolver.arpa(encrypted-DNS discovery) is forwarded upstream instead of being answered locally.
For developers
- New ADRs: 153–164. 157–160 (the dashboard and top-list ones) are still Proposed.
- New MCP tools:
list_anomaly_mutes,plan_set_anomaly_mute,plan_remove_anomaly_mute;find_anomaliestakesmuted;top_itemstakesfrom,to, andhour: rolling. - New API paths:
PUTandDELETE /api/v1/anomaly-mutes/{name},POST /api/v1/system/support-packageandGET …/{id}/map,DELETE /api/v1/cluster/nodes/{id}. New token scopeconfig:write:anomalies. The support package is admin-only, with no agent scope and no MCP tool. - New config sections:
[[anomaly_mute]](shared),[support]and[watchdog](node-local, restart-only). New metrics:telltale_anomalies_muted_total,telltale_watchdog_enabled,telltale_watchdog_check_age_seconds,telltale_watchdog_last_action_timestamp_seconds. - New crate
telltale-support(redaction, pseudonyms, the archive).telltale-netgainsstacks, the only newunsafe, with no new dependency. - CI: the watchdog end-to-end test runs a real process wedged on purpose (a
wedge-testCargo feature that no release build has).
Thanks
Greg Pakes for the field reports (#2–#7), the idea behind mutes (#38), the Top clients names (#8), and the dashboard pull requests; John Depon for stub zones.
What's Changed
- fix(api): Top clients names unnamed devices like the query log [CLU-002] by @gregpakes in #8
- Stub zones: resolve a domain from its own authoritative servers [DNS-022] by @JohnDepon in #10
- build(deps): Bump hono from 4.13.13 to 4.13.18 in /deploy/mcp-e2e by @dependabot[bot] in #13
- build(deps-dev): Bump vite from 8.3.2 to 8.3.8 in /ui by @dependabot[bot] in #14
- build(deps): bump @hono/node-server from 2.1.3 to 2.1.8 in /deploy/mcp-e2e by @dependabot[bot] in #25
- build(deps): bump qs from 6.16.0 to 6.17.1 in /deploy/mcp-e2e by @dependabot[bot] in #27
- build(deps): bump the ui-minor-patch group across 1 directory with 4 updates by @dependabot[bot] in #28
- build(deps): bump the rust-minor-patch group across 1 directory with 13 updates by @dependabot[bot] in #20
- build(deps): bump ed25519-dalek from 2.2.0 to 3.0.0 by @dependabot[bot] in #21
- build(deps): bump base64 from 0.22.1 to 0.23.1 by @dependabot[bot] in #22
- build(deps): bump rusqlite from 0.41.1 to 0.42.0 by @dependabot[bot] in #24
- build(deps): bump body-parser from 2.3.0 to 2.4.0 in /deploy/mcp-e2e by @dependabot[bot] in #32
- build(deps): bump zod from 4.6.5 to 4.6.7 in /deploy/mcp-e2e by @dependabot[bot] in #33
- build(deps): bump @modelcontextprotocol/sdk from 1.32.1 to 1.32.2 in /deploy/mcp-e2e in the mcp-e2e-minor-patch group across 1 directory by @dependabot[bot] in #15
- build(deps): bump fast-uri from 3.1.8 to 3.1.9 in /deploy/mcp-e2e by @dependabot[bot] in #26
- build(deps): bump minisign-verify from 0.2.5 to 0.3.0 by @dependabot[bot] in #47
- build(deps): bump ip-address from 10.7.3 to 10.7.6 in /deploy/mcp-e2e by @dependabot[bot] in #45
- build(deps): bump parking_lot from 0.12.5 to 0.13.0 by @dependabot[bot] in #46
- build(deps): bump rustls from 0.23.45 to 0.23.46 in /fuzz by @dependabot[bot] in #16
- build(ci): bump helm/chart-testing-action from 2.7.0 to 2.9.0 in the actions group across 1 directory by @dependabot[bot] in #48
- build(deps): bump prost from 0.14.5 to 0.14.6 in the rust-minor-patch group by @dependabot[bot] in #54
- build(ci): bump actions/setup-node from 4 to 7 by @dependabot[bot] in #53
- build(ci): bump docker/setup-qemu-action from 3 to 4 by @dependabot[bot] in #51
- build(ci): bump actions/upload-artifact from 4 to 7 by @dependabot[bot] in #52
New Contributors
- @gregpakes made their first contribution in #8
- @JohnDepon made their first contribution in #10
- @dependabot[bot] made their first contribution in #13
Full Changelog: v0.6.0...v0.7.0