Session — Apr 23 — IVR visual builder → prod + 11 staging fixes¶

Full ship day. IVR visual builder feature (previously staging-only) landed on prod via PR #63 after clearing 11 staging UAT bugs, 1 auth bug, and 1 hand- editing prod Asterisk config. Production orgs: Om Chamber, GrandEstancia, Zauto AI. Zero customer-reported incidents.

This page is the reference for the fixes + troubleshooting paths uncovered. For the prod ship runbook, see Prod Direct-Edit. For the complete test suite, see Test Cases.

Shipped to prod (PR #62 + PR #63)¶

PR #62 — Auth auto-logout fixes¶

PR #63 — IVR visual builder (14 commits) + UAT bug fixes¶

Ten feature commits for IVR Phase 1/2 + React Flow builder + misc UX, plus four session fixes:

Prod Asterisk hand-edit (not in monorepo)¶

One config file outside /opt/<app> was edited after explicit user authorization:

Backup saved at /etc/asterisk/ext_from_cloud.conf.bak-2026-04-23 on prod.

Troubleshooting reference (hard-won lessons)¶

"IVR call lands but no greeting plays — just hangs up"¶

Diagnostic: 1. asterisk -rx "dialplan show <ivr_ext>@<prefix>__ivr" — is there a Background(...) priority? 2. Check the filename passed: Background(greetings/<prompt>). Must have greetings/ prefix. Bare filename fails silently because Asterisk looks only in <astsoundsdir>/<lang>/, not in subdirs. 3. ls /var/lib/asterisk/sounds/greetings/<prompt>.wav — does the file exist?

Fix: - If path is missing greetings/ prefix: bug in dialplanGenerator.js:567 (fixed in PR #63). Regenerate the org's dialplan via configDeploymentService.deployOrganizationConfiguration. - If file missing: TTS was never generated. Either re-click "Generate greeting" in the IVR UI, or call TTSService directly:

const tts = new TTSService();
await tts.saveGreetingAudio(`ivr_${ivrId}`, text, language, voice);

"SIP phone dials IVR / queue extension → 404 Not Found"¶

Diagnostic:

asterisk -rx "dialplan show <number>@<prefix>__internal"

Fix paths:

"Outbound call returns 403 / 404 — pjsip endpoint unknown"¶

Diagnostic:

grep -iE "error|warn" /var/log/asterisk/full.log | grep <endpoint-name>

Common error: Could not find option suitable for category '...' named 'X' means pjsip rejected the endpoint config at load time due to an unknown option. The endpoint is then missing from runtime — any Dial(...@endpoint) returns congestion.

Most common cause (historically): sip_trunks.configuration column contains metadata keys (system_trunk, nuc_gateway, channels, etc.) that get splatted into the pjsip file. Fixed in PR #63 (sipTrunkService.js now filters these). After the fix is deployed, regenerate the org's config and asterisk -rx "pjsip reload".

"Admin impersonating — users don't load, no auto-logout"¶

Symptom: Admin impersonated an org 24h+ ago. Page shows empty users list and zero call stats. Not kicked to login.

Diagnostic:

// In browser devtools on editor.astradial.com
localStorage.getItem('pbx_org_token_exp')  // < Date.now() = expired
localStorage.getItem('gateway_admin_key')  // truthy = admin session active
JSON.parse(localStorage.getItem('org_access')).impersonating  // true

If all three are true and you're not being redirected, the watcher isn't firing — pre-PR-#62 behaviour. Fix landed in PR #62: ship to prod via CI/CD.

"Outbound dial with + prefix (E.164) doesn't match"¶

Zoiper and other softphones dial international numbers with a leading +. Asterisk _X. patterns do NOT match + (it's a literal char, not a digit).

Fix (PR #63): outbound context gets a _+X. catch-all that strips + and re-enters at ${EXTEN:1},1.

exten => _+X.,1,NoOp(Stripping leading + from ${EXTEN})
exten => _+X.,n,Goto(${EXTEN:1},1)

"Staging outbound hits prod but prod says 'congested' / 'number not in service'"¶

Flow: staging dials PJSIP/<num>@org_mna9x47k_tata → sends INVITE to prod's 10.10.10.1:5060 over WireGuard → prod matches endpoint cloud-endpoint-stage (by IP) → enters [from-cloud] context.

If prod's ext_from_cloud.conf Gotos a context that doesn't exist, the call dies. As of 2026-04-23 it Gotos staging-outbound (defined in ext_staging_outbound.conf) which dials tata_gateway. If that file is reverted or Gotos a per-org context for an org not provisioned on prod, staging outbound stops working. Check:

cat /etc/asterisk/ext_from_cloud.conf
dialplan show staging-outbound

"Dialplan regen works on one org but fails on another with 'Unknown column'"¶

Check for schema drift. Staging has gone through more migration-like ALTERs than prod. Example seen today: prod ivrs table was missing greeting_text / greeting_language / greeting_voice columns because the idempotent ALTERs in the migration never ran on prod.

Fix:

scp api/database/migrations/<migration>.sql root@<vps>:/tmp/
mariadb -u<user> -p<pw> pbx_api_db < /tmp/<migration>.sql

"Om Chamber queue 5001 call silently goes to PSTN"¶

This was a prod bug revealed by the include-order fix. Before PR #63, dialling queue number 5001 from an internal SIP phone on Om Chamber matched _X. in the outbound context (because _outbound was included first). The queue never received the call. After the fix, exact 5001 in _queue wins.

Verification after any dialplan change¶

Minimum verification after regenerating or editing dialplan:

# 1. Reload cleanly
asterisk -rx "dialplan reload" 2>&1 | tail
asterisk -rx "pjsip reload" 2>&1 | tail

# 2. Spot-check each routing type
asterisk -rx "dialplan show <queue>@<org>_internal"          # should hit _queue
asterisk -rx "dialplan show <ivr_ext>@<org>_internal"        # should hit _ivr
asterisk -rx "dialplan show <pstn_number>@<org>_internal"    # should hit _outbound _X.
asterisk -rx "dialplan show _+X.@<org>_outbound"             # should exist
asterisk -rx "pjsip show endpoint <trunk-peer-name>"         # should be loaded, not "Unable to find object"

# 3. Tail logs for errors
tail -100 /var/log/asterisk/full.log | grep -iE "error|warning" | grep -v "tata_gateway/tag"

# 4. A live PSTN call — irreplaceable

Deployment timeline (for audit)¶

All times IST. Prod = 89.116.31.109, Staging = 94.136.188.221.

Related docs¶

• Test Cases (full suite) — run these against any future IVR / dialplan change.
• Prod Direct-Edit — the runbook for working on prod VPS directly.
• Staging Direct-Edit — same idea for staging.
• Staging Environment — architecture.