Skip to content

Archive ​

Validators prune. Substrate default --state-pruning is 256 blocks ≈ 25.6 minutes at 6 s. state_getStorage(…, hash#old) on rpc2 returns UnknownBlock: State already discarded once the window slides.

An archive node keeps every block and every state:

--state-pruning archive --blocks-pruning archive

What is deployed ​

An archive unit runs on validator 2 (212.227.108.41), systemd empooriochain-archive: P2P 8449, RPC 9948. Measured 2026-09-05: full sync in minutes (~437 blocks/s), about 1.3 GB. At ~60k blocks, a from-scratch archive was 129 s and 1.0 GB. That cost grows without prune, by design.

Proof with block 1000:

Nodestate_getStorage(…, hash#1000)
Archiveresult (including null for empty)
ValidatorUnknownBlock: State already discarded

Same machine as a validator ​

Validator 2 had more free disk. Sharing a machine is a compromise: archive I/O could delay authoring, and with two validators a slow one stops finality. The unit is capped (CPUQuota=150%, MemoryMax=2G, Nice=10) so the validator always wins. During the first sync, load sat at 1.11 / 4 CPU and the validator kept finalizing.

The right medium-term shape is a third machine that also serves public RPC, so validators can drop --unsafe-rpc-external.

Disk ​

An archive grows forever. Watch free disk before it is urgent. Public RPC rpc2 is the validator endpoint, not the archive. Historical state queries against rpc2 will fail outside the prune window. EmpooScan needs an archive (or a similar store) for old state.

English technical docs. Marketing lives at empooriochain.com.