cscott/ask-annie
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Add chapters for session 813464822 (Haiku)

Browse Source
This commit is contained in:
Conan Scott
2026-03-24 10:02:32 +00:00
parent c3461c6441
commit 3cc83394f9
1 changed files with 20 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

20
chapters/813464822.json Normal file
View File

@@ -0,0 +1,20 @@
[
{"timestamp": "0:19", "title": "PGP Key Expiration and Route Selection", "summary": "When a PGP key expires and is replaced with a new key of identical name, the route may not automatically recognize the updated key and requires manual re-selection. Routes reference keys by name; possible cause is a caching issue where the route caches the old key reference. Recommend opening a support ticket if this occurs."},
{"timestamp": "1:10", "title": "Key Naming Best Practices", "summary": "Recommended format: PARTNER_NAME_CURR_PGP (uppercase partner name, underscore, 'curr' for current, underscore, 'pgp'). Avoid case-sensitivity mismatches. Note whether delete-and-recreate vs replace was used when updating keys in support tickets."},
{"timestamp": "2:13", "title": "Key Lookup Mechanism and Cache Behavior", "summary": "Routes reference keys by name, not by ID. Key lookups happen during route execution only. Server restart may resolve cache-related key selection failures where newly imported key under same name is not recognized."},
{"timestamp": "6:40", "title": "Certificate Expiration Monitoring in ST", "summary": "Account Maintenance application monitors certificate expiration for all account-level certificates: login, BGP, and SSH-created certificates. Notification days are configurable per certificate type. Global certificates are NOT monitored by this application and require separate manual monitoring."},
{"timestamp": "9:48", "title": "Selective Monitoring for PGP Keys Only", "summary": "To use Account Maintenance exclusively for PGP key monitoring, set account certificate notification threshold to an extremely high future date (e.g., 10,000 days) to effectively disable account-level notifications while keeping key monitoring active."},
{"timestamp": "11:01", "title": "Locating Key Usage Across Routes", "summary": "Hard-coded name references are searchable via XML export of routes. Expression-based key references (via variables/parameters) are not discoverable via search. Submit an ideas portal request for a lookup feature if needed."},
{"timestamp": "12:32", "title": "Certificate ID Lookup via REST API", "summary": "REST API supports filtering certificates by ID. Use GET certificate endpoints with ID query parameter to identify which certificate matches a partner-provided key ID. Direct ID filtering is the correct approach; brute-force enumeration of all keys is inefficient."},
{"timestamp": "17:01", "title": "Transfer Site Usage Across Routes", "summary": "Identifying all routes using a transfer site requires XML export of routes (not accounts, to avoid data bloat). Search exported XML for the site name. Expression-language site references are not discoverable via search."},
{"timestamp": "20:15", "title": "Cluster Types and Downtime Upgrade Strategy", "summary": "Standard cluster has no zero-downtime upgrade planned; always requires downtime. Enterprise cluster zero-downtime upgrades (ST-I-140) planned, coming to cloud first. Standard cluster's single-database model does not support zero-downtime upgrades."},
{"timestamp": "26:03", "title": "Breaking Cluster for Minimal Downtime", "summary": "Breaking a Standard cluster creates lower (not zero) downtime but carries data loss risk. Any simultaneous two-way transfers can lose history during the split. Changes made on non-primary side during split are lost during resync. Manual file resubmission required for lost records."},
{"timestamp": "32:46", "title": "Enterprise Cluster Node Offline and Sync Issues", "summary": "Enterprise clusters use a single database; no explicit sync between nodes. Cache synchronization issues cause node offline appearances. Cause: misconfigured cache, database latency, or coherence port (8005) communication timeout. Annie suspects post-5.5-upgrade regression."},
{"timestamp": "34:25", "title": "Enterprise Cluster Cache Architecture", "summary": "Both enterprise nodes reference same database but maintain separate local caches. Cache updates should be instantaneous. If not: configuration issue, database latency, or blocked cache replication. Check capacity guide for cache configuration; if standard config untouched, environment issue is present."},
{"timestamp": "39:31", "title": "Cluster Time Sync and NTP Configuration", "summary": "Clock skew between cluster nodes causes cache confusion and transaction delays. Linux clusters must use NTP. Example: update at 9:00 AM on node 1, node 2 thinks it's 8:59 AM, causing replication delays. Verify time sync between nodes; misconfigured NTP is a common cause of cluster issues."},
{"timestamp": "40:34", "title": "Split Cluster Node Behavior During Outage", "summary": "When one cluster node marks the other offline, both nodes independently decide which jobs to run and may have different cache states. Same transfer could behave differently on node 1 vs node 2. Root cause must be identified to avoid cascading failures."},
{"timestamp": "41:41", "title": "Route Export and Import Limitations", "summary": "No built-in route export/import feature exists. Full account export/import works; partial account export (subset of subscriptions, routes, keys) is not supported. Moving partial account configuration between instances (the 'promotion' problem) remains unsolved."},
{"timestamp": "44:22", "title": "Notification Filtering by Protocol Direction", "summary": "session.protocol variable distinguishes initial uploads (HTTP, SSH) from routing republishes ('routing'). Use route condition 'not session.protocol equals routing' to filter out internal republishes. Place notifications on individual routes, not package level."},
{"timestamp": "45:45", "title": "Filtering Inbound vs Outbound Transfer Notifications", "summary": "When a route unzips and republishes files, subsequent route sees the file as 'routing' protocol. Build two separate routes: one for initial upload (sends notifications), one for routing republish (skips notifications). Use session.protocol condition on each route separately."},
{"timestamp": "49:49", "title": "Production Disaster Recovery Patching Strategy", "summary": "For Enterprise cluster on Oracle: DBA must sever database connection between prod and DR before patching either, maintaining independent databases during upgrade. After DR upgrade, reconnect and resync; one database overwrites the other. All changes on non-primary side are lost during resync."}
]