Перейти к основному содержимому

Restarting a Solana Cluster

Step 1. Identify the latest optimistically confirmed slot for the cluster​

In Agave 1.14 or greater, run the following command to output the latest optimistically confirmed slot your validator observed:

agave-ledger-tool -l ledger latest-optimistic-slots

In Agave 1.13 or less, the latest optimistically confirmed can be found by looking for the more recent occurrence of this metrics datapoint.

Call this slot SLOT_X

Note that it's possible that some validators observed an optimistically confirmed slot that's greater than others before the outage. Survey the other validators on the cluster to ensure that a greater optimistically confirmed slot does not exist before proceeding. If a greater slot value is found use it instead.

Step 2. Stop the validator(s)​

Step 3. Optionally install the new solana version​

Step 4. Create a new snapshot for slot SLOT_X with a hard fork at slot SLOT_X​

$ agave-ledger-tool -l <LEDGER_PATH> --snapshot-archive-path <SNAPSHOTS_PATH> --incremental-snapshot-archive-path <INCREMENTAL_SNAPSHOTS_PATH> create-snapshot SLOT_X <SNAPSHOTS_PATH> --hard-fork SLOT_X

The snapshots directory should now contain the new snapshot. agave-ledger-tool create-snapshot will also output the new shred version, and bank hash value, call this NEW_SHRED_VERSION and NEW_BANK_HASH respectively.

Adjust your validator's arguments:

 --wait-for-supermajority SLOT_X
 --expected-bank-hash NEW_BANK_HASH

Then restart the validator.

Confirm with the log that the validator booted and is now in a holding pattern at SLOT_X, waiting for a super majority.

Once NEW_SHRED_VERSION is determined, nudge foundation entrypoint operators to update entrypoints.

Step 5. Announce the restart on Discord:​

Post something like the following to #announcements (adjusting the text as appropriate):

Hi @Validators,

We've released v1.1.12 and are ready to get testnet back up again.

Steps:

  1. Install the v1.1.12 release: https://github.com/solana-labs/solana/releases/tag/v1.1.12
  2. a. Preferred method, start from your local ledger with:
agave-validator
  --wait-for-supermajority SLOT_X     # <-- NEW! IMPORTANT! REMOVE AFTER THIS RESTART
  --expected-bank-hash NEW_BANK_HASH  # <-- NEW! IMPORTANT! REMOVE AFTER THIS RESTART
  --hard-fork SLOT_X                  # <-- NEW! IMPORTANT! REMOVE AFTER THIS RESTART
  --no-snapshot-fetch                 # <-- NEW! IMPORTANT! REMOVE AFTER THIS RESTART
  --entrypoint entrypoint.testnet.solana.com:8001
  --known-validator 5D1fNXzvv5NjV1ysLjirC4WY92RNsVH18vjmcszZd8on
  --expected-genesis-hash 4uhcVJyU9pJkvQyS88uRDiswHXSCkY3zQawwpjk2NsNY
  --only-known-rpc
  --limit-ledger-size
  ...                                # <-- your other --identity/--vote-account/etc arguments

b. If your validator doesn't have ledger up to slot SLOT_X or if you have deleted your ledger, have it instead download a snapshot with:

agave-validator
  --wait-for-supermajority SLOT_X     # <-- NEW! IMPORTANT! REMOVE AFTER THIS RESTART
  --expected-bank-hash NEW_BANK_HASH  # <-- NEW! IMPORTANT! REMOVE AFTER THIS RESTART
  --entrypoint entrypoint.testnet.solana.com:8001
  --known-validator 5D1fNXzvv5NjV1ysLjirC4WY92RNsVH18vjmcszZd8on
  --expected-genesis-hash 4uhcVJyU9pJkvQyS88uRDiswHXSCkY3zQawwpjk2NsNY
  --only-known-rpc
  --limit-ledger-size
  ...                                # <-- your other --identity/--vote-account/etc arguments
 You can check for which slots your ledger has with: `agave-ledger-tool -l path/to/ledger bounds`
  1. Wait until 80% of the stake comes online

To confirm your restarted validator is correctly waiting for the 80%: a. Look for N% of active stake visible in gossip log messages b. Ask it over RPC what slot it's on: solana --url http://127.0.0.1:8899 slot. It should return SLOT_X until we get to 80% stake

Thanks!

Step 7. Wait and listen​

Monitor the validators as they restart. Answer questions, help folks,

Troubleshooting​

80% of the stake didn't participate in the restart, now what?​

If less than 80% of the stake join the restart after a reasonable amount of time, it will be necessary to retry the restart attempt with the stake from the non-responsive validators removed.

The community should identify and come to social consensus on the set of non-responsive validators. Then all participating validators return to Step 4 and create a new snapshot with additional --destake-vote-account <PUBKEY> arguments for each of the non-responsive validator's vote account address

$ agave-ledger-tool -l ledger create-snapshot SLOT_X ledger --hard-fork SLOT_X \
    --destake-vote-account <VOTE_ACCOUNT_1> \
    --destake-vote-account <VOTE_ACCOUNT_2> \
    .
    .
    --destake-vote-account <VOTE_ACCOUNT_N> \

This will cause all stake associated with the non-responsive validators to be immediately deactivated. All their stakers will need to re-delegate their stake once the cluster restart is successful.