Install troubleshooting

When the installer
doesn't go quietly.

Common failure modes during install or first launch, with the recovery for each. If your error isn't covered here, the platform's own logs are usually the next port of call — journalctl -u eldric-aios --since "10 min ago" for the daemon, and /var/log/eldric-install.log for the bootstrap installer and eldric setup output.

1. curl: command not found

The base distribution doesn't have curl. Install it first, then re-run the bootstrap:

sudo dnf install curl
curl -fsSL https://repo.eldric.ai/install.sh | sudo bash

2. GPG fingerprint mismatch or GPG check FAILED

The bootstrap script pins the Eldric signing key by fingerprint; if the fetched key doesn't match the pinned value, the script refuses to proceed. That's the safe behaviour — it means an intermediary served a swapped key, or the local /etc/pki/rpm-gpg/RPM-GPG-KEY-eldric is stale.

Remove any stale local key and re-run the bootstrap; it will fetch a fresh copy and verify the fingerprint:

sudo rm -f /etc/pki/rpm-gpg/RPM-GPG-KEY-eldric
curl -fsSL https://repo.eldric.ai/install.sh | sudo bash

If the fingerprint check still fails, your network is intercepting HTTPS — investigate the corporate proxy or HTTPS-inspection appliance before retrying.

3. No package eldric-aios available

The repository file wasn't installed, or dnf didn't see it. Confirm the repo is registered:

sudo dnf repolist | grep eldric

If nothing matches, re-run the bootstrap script. If the repo is listed but the package isn't found, your dnf cache is stale: sudo dnf clean all && sudo dnf makecache.

4. /var/lib/eldric already exists — refusing first-time bootstrap

The bootstrap installer (install-eldric.sh) refuses to run if a prior install left state behind — that protects existing data from being overwritten by a first-run path. If this is genuinely an upgrade, run sudo dnf upgrade eldric-aios instead. If the previous install is unwanted and you have nothing to preserve, archive /var/lib/eldric and /data/eldric elsewhere first, then retry.

5. Port 8880 already in use or Port 443 already in use

Another service is bound to the port Eldric wants — most often nginx or Apache on 443, or a previous Eldric run still holding 8880. Two options:

• Stop the conflicting service if it's no longer needed: sudo systemctl stop nginx && sudo systemctl disable nginx
• Change Eldric's port by editing /etc/eldric/eldric-aios.env (single-node default is ELDRIC_AIOS_PORT=8880; edge-fronted clusters use ELDRIC_EDGE_HTTPS_PORT=443). Restart: sudo systemctl restart eldric-aios.

6. Cannot open /chat — connection refused

The daemons need a few seconds to come up. Single-node listens on http://<host>:8880/chat by default (the URL drops the port only when a dedicated edge gateway is in front). Wait thirty seconds after install, then check:

sudo systemctl status eldric-aios
journalctl -u eldric-aios --since "5 min ago" --no-pager | tail -40

If the unit is active (running) but the port is silent, a firewall is most likely blocking it: sudo firewall-cmd --add-port=8880/tcp --permanent && sudo firewall-cmd --reload. Re-running sudo eldric setup prints a summary that includes the actual port and the health-probe result, which is usually faster than reading the unit logs.

7. eldric setup times out waiting for /health

The post-install probe waits up to sixty seconds for the local daemon's /health endpoint to come green. If it times out, the daemon is up but not ready — usually a slow first-boot model warm-up, or a configuration error that's keeping a module from initialising. Two checks:

journalctl -u eldric-aios --since "2 min ago" --no-pager | tail -40
sudo eldric setup --json

eldric setup is idempotent and safe to re-run; the --json flag emits a machine-readable summary that names which probe failed. The full bootstrap log lives at /var/log/eldric-install.log.

8. Admin signup says “an admin already exists”

Someone on the network signed up first. The admin role is locked to the first signup on a fresh cluster. Three paths:

• Have the existing admin add you from Admin Console → Users.
• If the wrong person claimed admin by accident, the existing admin can reassign the role to you and downgrade themselves.
• Last resort on a clean install you control: stop the service, wipe /var/lib/eldric/controller/users.db, restart, sign up again. This destroys all users and tenants; never do it on a production cluster.

9. License activation: invalid signature

The license file is either corrupted in transit or older than the embedded public key expects. Re-request the latest signed file from license@core.at. Don't edit the JSON by hand — any modification invalidates the Ed25519 signature. To retry activation from the command line:

sudo eldric setup --license-file /path/to/license.json

10. macOS: “Eldric.pkg cannot be opened — unidentified developer”

macOS Gatekeeper guards against unsigned packages. Eldric 5.0 PKGs are not Apple-notarized (notarized releases are planned for an upcoming 5.0.x patch). To bypass:

• Right-click Eldric-latest-macos.pkg in Finder.
• Choose Open from the context menu.
• Confirm with Open in the security dialog.

11. Where to get more help.

The platform produces detailed logs that almost always have the answer:

journalctl -u eldric-aios --since today --no-pager
cat /var/log/eldric-install.log

If the logs aren't enough, write to office@eldric.ai. Paid-tier customers reach support@eldric.ai with their license ID for prioritised handling.