Bug bash 2: 11 bugs in Stacks, Alerts, Docker, skins and the installer, fixed and verified (BUGS.md)

Stacks: deploy with every compose file (labels from inspect, not comma-joined ps);
refuse a stale overwrite; empty list without Docker. Alerts: firing state survives a
restart without re-announcing; vanished alerts are recorded as cleared with the reason;
rule values validated; 200 rules. Docker: 404 for a removed container's logs. Skins: four
palettes lifted to readable contrast. Phone: bigger toggles and swatches. Installer:
deploy republishes; an older published build is never installed silently.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
This commit is contained in:
Mitchell-Johnson
2026-09-19 15:59:51 -07:00
co-authored by Claude Fable 5.1
parent 9c4f9dfa9f
commit eddba4250b
10 changed files with 320 additions and 68 deletions
+74 -3
View File
@@ -71,8 +71,18 @@ type alertSaved struct {
Channels []AlertChannel `json:"channels"`
Events []AlertEvent `json:"events"`
Seeded bool `json:"seeded"`
// Firing is what was firing when the hub last looked, so a restart (every deploy is one)
// carries on where it was and doesn't announce the same alerts all over again.
Firing map[string]alertKept `json:"firing,omitempty"`
}
type alertKept struct {
Since int64 `json:"since"`
Value any `json:"value"`
}
const maxRules = 200
type alertState struct {
pending time.Time
firing bool
@@ -109,6 +119,9 @@ func newAlertEngine(s *Server) *alertEngine {
if !e.saved.Seeded {
e.saved.Rules, e.saved.Seeded = append([]AlertRule{}, defaultRules...), true
}
for key, k := range e.saved.Firing {
e.state[key] = &alertState{pending: time.UnixMilli(k.Since), firing: true, since: time.UnixMilli(k.Since), value: k.Value}
}
for i, r := range e.saved.Rules { // the first day's default watched "not running", which a container stopped on purpose also is
if r.ID == "container-down" && r.Sensor == "container.*.running" {
e.saved.Rules[i].Name, e.saved.Rules[i].Sensor, e.saved.Rules[i].Value = "Container down", "container.*.down", true
@@ -275,12 +288,47 @@ func (e *alertEngine) evaluate() {
}
}
}
for key := range e.state { // a sensor that is gone (a removed container, a deleted rule) takes its state with it
if !seen[key] {
delete(e.state, key)
byID := map[string]AlertRule{}
for _, r := range rules {
byID[r.ID] = r
}
for key, st := range e.state { // no longer watched: the rule was paused, deleted or narrowed, or the sensor is gone
if seen[key] {
continue
}
if st.firing { // it was on the Alerts tab and perhaps on a phone: say that it is over, and why
part := strings.SplitN(key, "|", 3)
r, known := byID[part[0]]
name, why := r.Name, "no longer watched"
if !known {
name, why = part[0], "rule deleted"
} else if !r.Enabled {
why = "rule paused"
}
if name == "" {
name = part[0]
}
events = append(events, AlertEvent{T: now.UnixMilli(), Kind: "resolved", Rule: name, RuleID: part[0], Machine: part[1], Sensor: part[2], Value: why, Severity: r.Severity})
}
delete(e.state, key)
}
e.last = now
kept := map[string]alertKept{}
for key, st := range e.state {
if st.firing {
kept[key] = alertKept{Since: st.since.UnixMilli(), Value: st.value}
}
}
changed := len(kept) != len(e.saved.Firing)
for k := range kept {
if _, ok := e.saved.Firing[k]; !ok {
changed = true
}
}
e.saved.Firing = kept
if changed && len(events) == 0 {
e.persist()
}
if len(events) > 0 {
e.saved.Events = append(e.saved.Events, events...)
if len(e.saved.Events) > maxEvents {
@@ -390,6 +438,10 @@ func (s *Server) alertsPut(w http.ResponseWriter, r *http.Request) {
jsonErr(w, 400, "bad request")
return
}
if body.Rules != nil && len(*body.Rules) > maxRules {
jsonErr(w, 400, fmt.Sprintf("at most %d rules: every one is checked on every machine every %d s", maxRules, int(alertEvery.Seconds())))
return
}
if body.Rules != nil {
ids := map[string]bool{}
for i := range *body.Rules {
@@ -408,6 +460,25 @@ func (s *Server) alertsPut(w http.ResponseWriter, r *http.Request) {
jsonErr(w, 400, fmt.Sprintf("rule %q: the comparison must be one of > >= < <= == !=", ru.Name))
return
}
switch val := ru.Value.(type) {
case float64, bool:
case string:
if len(val) > 200 {
jsonErr(w, 400, fmt.Sprintf("rule %q: the value is too long", ru.Name))
return
}
if ru.Op != "==" && ru.Op != "!=" { // "cpu.usage > ninety" can never be true, and would never say so
jsonErr(w, 400, fmt.Sprintf("rule %q: %s compares numbers, and %q is not one; text can only be == or !=", ru.Name, ru.Op, val))
return
}
default:
jsonErr(w, 400, fmt.Sprintf("rule %q: the value must be a number, true/false or text", ru.Name))
return
}
if _, isBool := ru.Value.(bool); isBool && ru.Op != "==" && ru.Op != "!=" {
jsonErr(w, 400, fmt.Sprintf("rule %q: true/false can only be == or !=", ru.Name))
return
}
if ru.Severity != "info" && ru.Severity != "crit" {
ru.Severity = "warn"
}