Skip to main content
All Injective mainnet node operators must upgrade ASAP to prevent an AppHash mismatch. New consensus rules are enforced after height 191,016,661 on injective-1. This is a binary swap and restart: there is no upgrade handler. Cosmovisor will not activate this release from a new upgrade directory.

​
Context

On November 9, Injective carried out a mainnet upgrade as part of its network readiness program. The program tests emergency upgrade procedures and coordination across validators and ecosystem partners through periodic, unannounced exercises. The exercise completed successfully and normal network operations resumed. Operators still running an older binary must upgrade their nodes.

​
Required action

This applies to validators, sentries, RPC and broadcast nodes, and nodes supporting exchanges, custodians, market makers, indexers, explorers, relayers, and other integrations. If you use a third-party node provider, confirm they have upgraded.

​
1. Download and verify

Official release · Release manifest Download into a staging directory and compare the archive’s SHA-256 against its entry in the checksum file before extracting or installing. Use sha256sum on Linux or shasum -a 256 on macOS. Docker: use injectivelabs/injective-core:v1.20.4-bl.3. Verify without mounting node data:
Expected version:

​
2. Stop and back up

  1. Stop your node service, Cosmovisor, or container deployment. Confirm injectived has exited and its supervisor cannot restart it during the swap. A halted chain does not mean the node process has stopped.
  2. Back up the existing binary and node configuration. Preserve existing node data; a normal binary swap does not require a snapshot restore.
  3. Validators: after shutdown, back up <node-home>/data/priv_validator_state.json outside the data directory. Remote signer operators must preserve their signer’s state using their existing procedure.
Before any snapshot recovery, also back up node data and keys. Use your configured node home; ~/.injectived is only the default.

​
3. Swap the binary

​
Managed binary swap without Cosmovisor

For a systemd-managed node, use your actual unit name (the example is injectived). Check ExecStart to identify the executable, node home, user, and startup flags:
Confirm the process has exited and complete the backups in step 2 before replacing any files.
  1. Replace the injectived executable your service actually uses with the verified release binary.
  2. Preserve ownership and executable permissions. On Linux, ensure the installed executable has its execute bit set (chmod +x <installed-binary-path>).
  3. Install the bundled libwasmvm.x86_64.so where your Linux AMD64 service loads it. The official release uses /usr/lib and runs sudo ldconfig. Verify that location’s library:
Expected SHA-256:
If your deployment loads the library from another location, verify that file instead. If it already matches, replacement is unnecessary. Do not start with a mismatched library. After verifying the installed version and runtime library, restart through the same service manager:
Do not launch a second node manually alongside the managed service. Continue with the health checks in step 4.

​
Managed binary swap with Cosmovisor

Stop the systemd unit that runs Cosmovisor, even if that unit is named injectived. Use your actual unit name in place of cosmovisor below:
Read DAEMON_HOME from the service configuration, including any environment file. It may not be set in your interactive shell. Complete the backups in step 2 after shutdown.
  1. With Cosmovisor and its child process stopped, replace $DAEMON_HOME/cosmovisor/current/bin/injectived with the verified binary.
  2. Preserve ownership and executable permissions, and verify the runtime library as above.
  3. Do not create a new upgrades/<name> directory. There is no upgrade handler to switch binaries automatically.
Verify the exact active binary, then start the same Cosmovisor unit:
Confirm Cosmovisor starts the expected binary and continue with the health checks in step 4.

​
Docker or another container deployment

Update your existing deployment to injectivelabs/injective-core:v1.20.4-bl.3, preserving its data volumes, configuration, user, and startup arguments. The image includes runtime libraries. For Docker Compose, change the node service’s image, then pull and recreate it using your existing Compose configuration. Set node_service to your actual service name:
Only recreate the service once required recovery is complete and validator signing state is intact. Restarting the old container alone does not apply an image change.

​
4. Verify and restart ASAP

Run version against the exact installed executable or image and confirm v1.20.4-bl.3 (e0a1168). Cosmovisor operators should run:
Restart your service or Cosmovisor ASAP. Confirm:
  • The node is on injective-1, block height advances, and it catches up to the network.
  • Logs show no AppHash mismatch or missing-library errors.
  • Validators participate normally; partner integrations receive current blocks.
Repeat for all sentries, RPC nodes, and other node infrastructure. Report in your coordination channel:

​
5. AppHash mismatch recovery

  1. Stop the node and its supervisor again. Confirm the process has exited.
  2. Preserve the latest validator signing state after shutdown, outside the data directory. Keep a backup of node data, keys, and config before restoring.
  3. Obtain the recovery files from the pre-upgrade snapshot manifest for 191,016,661. The JSON is a manifest, not the snapshot archive itself. Confirm database-backend compatibility and verify the snapshot against published checksums.
  4. Restore using the Chain Portal snapshot instructions and validator recovery precautions. Use the patched binary for replay.
  5. Restore your preserved validator signing state before restarting, or follow your remote signer’s recovery procedure. Never reset signing state, replace it with older state, or run two instances with the same validator key.
  6. Restart and repeat the health checks above.
Do not apply rollback commands from the older bl.2 incident to this release without explicit chain-team guidance.

​
Support

If recovery fails, keep the node stopped and share your version, height, database backend, and relevant logs in your existing Injective coordination channel or contact the team on Discord. Never share private keys or signer credentials.
Last modified on November 9, 2026