Monitoring, backup and upgrades¶
Monitoring¶
pqwerai-exporter serves Prometheus metrics on 127.0.0.1:9545/metrics. Scrape over an SSH tunnel or VPN, or allow
one monitoring IP (ufw.sh --metrics-from <IP>). Shipped alert rules (monitoring/alerts.yml):
| Alert | Meaning |
|---|---|
PqwChainStalled |
Head not advancing |
PqwFinalityStalled, PqwFinalityLagHigh |
Finalized checkpoint not advancing / far behind head |
PqwPeerIsolation |
Too few peers |
PqwSignerOffline |
Validator signer not voting |
PqwPersistentFork, PqwReorgDeep, PqwWrongChain |
Fork or identity problems |
PqwIssuanceMismatch |
Issuance counters disagree with expectations |
PqwDiskNearlyFull, PqwMemoryLow, PqwRpcErrorSpike |
Host health |
PqwAIProviderDown |
Optional AI provider service down |
Backup and restore¶
sudo /opt/pqwerai/current/share/vps/bin/pqwerai-backup create /srv/backups # stops, archives, restarts
sudo /opt/pqwerai/current/share/vps/bin/pqwerai-backup restore /srv/backups/pqwerai-core-<ts>-<genesis>.tar.zst
Restore verifies the checksum and that the archive belongs to this network's genesis. Validator signer journals are deliberately excluded — restoring a stale journal can cause a slashable double vote.
Upgrade and rollback¶
sudo pqwerai-release install pqwerai-v8-testnet-rc2-linux-amd64.tar.gz <published sha256>
sudo pqwerai-release activate <new revision> # restarts, health-checks, auto-rolls back on failure
sudo pqwerai-release rollback # previous release
pqwerai-release list ; pqwerai-release current
The datadir is never modified by the release tool. A release that changes protocol_version needs a coordinated,
announced activation height — read the release notes first.
Sources at 72057fe3ec: testnet/release/vps/INSTALL.md · testnet/release/monitoring/alerts.yml
Found a problem? Report a documentation issue · Build testnet-rc1+72057fe3ec