Latest

Ethereum

Glamsterdam postpones its arrival on Hoodi and halts the 200M gas limit

The Ethereum Foundation delays the activation of Glamsterdam on Hoodi, scheduled for October 26, over security concerns about the new gas calculation.

Glamsterdam postpones its arrival on Hoodi and halts the 200M gas limit
Glamsterdam postpones its arrival on Hoodi and halts the 200M gas limit

Ethereum's schedule moved four days before the date. The Ethereum Foundation set October 16 for the activation of Glamsterdam on Hoodi for October 26, 2026 at 17:42:48 UTC, at epoch 132,352 and slot 4,235,264, with a gas limit increase to 200 million in that same epoch. This Thursday, October 22, the foundation published a note that undoes the plan: the activation is postponed and there is no new date in the postponement announcement.

The reason is not a client that fails to start. Core developers decided on the ACDE #248 call, held today at 14:00 UTC, to delay the deployment to investigate how the new state-creation gas calculation (EIP-8037) interacts with contracts already deployed on mainnet, after ChainSecurity warned that several gas checks no longer protect what they used to protect.

Anyone running a node on Hoodi can no longer trust the October 16 version table. Anyone following the roadmap can see where Glamsterdam's real friction lies, which is not consensus or block size, but the price of state. Mainnet still has no date.

The date that was on the calendar

The October 16 announcement was unusually precise for a testnet. It set epoch 132,352, slot 4,235,264 and the exact time with a UNIX timestamp, and added that Hoodi would raise the gas limit to 200 million through the EIP-8261 schedule, just like Sepolia. The table closed with mainnet in three empty cells. The postponement removes the Amsterdam activation on execution, the Gloas one on consensus and that gas schedule.

Network Epoch Slot UTC time
Sepolia 353.024 11.296.768 October 6, 2026, 13:53:36
Hoodi (annulled) 132.352 4.235.264 October 26, 2026, 17:42:48
Mainnet undecided undecided undecided

Sepolia already went through that step on October 6 and has since been operating with 200 million gas per block, close to triple the previous budget. In the chronicle of that trial, the blocks reviewed moved between 52 and 92 million gas, from 26% to 46% of capacity.

Hoodi was going to repeat the move four days after today. The postponement leaves that second opinion up in the air, with clients already published and network configurations pointing to an epoch that no longer exists.

What did ChainSecurity find in the new gas calculation?

The auditor's analysis, based on the October 13 specifications, focuses on two proposals in the same package. EIP-8037 raises and separates the cost of creating state, while EIP-8038 updates the cost of accessing the technical report. The conclusion is that gas stops being a number that only goes down.

The new mechanism is a state gas reservoir. Everything a transaction puts above 16,777,216 gas goes in there, and state creation draws first from that reservoir before touching execution gas. Creating a slot from scratch costs 97,920 state gas, 64 bytes times 1,530 gas, and the execution charge drops to 10,100 warm and 12,100 cold versus the 20,000 and 22,100 of Osaka.

The practical consequence is what raised the alarms. If the reservoir pays the expensive part, a contract may be left with enough execution gas to finish an error path that previously ran out of gas and reverted. Contracts that infer from `gasleft()` whether a call failed due to lack of gas can no longer infer it, and an operation that should have been retryable is recorded as a definitive failure.

Serious risks

ChainSecurity sums it up this way: the gas a contract has left can go up. It does so when clearing a slot created in the same transaction, because the refund arrives as an immediate top-up of up to 97,920 units. The comment signed by pcaversaccio in the call thread speaks of serious risks with an impact on the application layer and warns that disclosure to the affected protocols is still ongoing.

That is the point that explains the decision. Glamsterdam does not fall because of a client that fails to start; it is postponed because of contracts already in production that have not yet been able to be reviewed under the affected label.

The package is not limited to gas. Glamsterdam brings into consensus the separation between who proposes the block and who builds it (EIP-7732), and adds per-block access lists that allow reading state in parallel before validating (EIP-7928).

What must anyone running a node on Hoodi do?

The first reflex, updating to the version in the October 16 table, is no longer valid. The postponement note says it plainly: those versions do not confirm that the schedule change is included, and configurations that still schedule Glamsterdam at the old time must be corrected. For now, the revised list included only two versions, Lodestar v1.50.1 on consensus and Besu 26.10.1 on execution.

For an operator, the order of checks is short and cannot be left for Monday, because a configuration that still points to the annulled epoch may start with rules no one else applies:

  • Update the execution client and the consensus client, and if you validate, the validator client too, because all three have to speak the same language.
  • Read your team's notes instead of the table from the previous announcement, and wait for a version that says the October 26 activation is out.
  • Check the Hoodi network configuration if you maintain it by hand, because epoch 132,352 is no longer the right one.

The rest of the October 16 list is still useful for knowing who is working on the fork, with six consensus implementations and six execution ones, and with Grandine still in release candidate. But none of those rows promises the new schedule, and the foundation itself warns that the table will be updated.

There is one validator detail that already bit on Sepolia. Prysm 7.2.0 and Teku 26.9.1 started there with 60 million gas by default after the fork, and anyone who wanted to propose with 200 million had to configure it by hand. That adjustment is back on Hoodi's to-do list.

Mainnet still has no date and gas marks the path

Mainnet still has no day. The postponement note repeats it twice: the Hoodi announcement does not set a mainnet date, and the upgrade changes nothing for users or ether holders. Today's call agenda included a notice with three candidates, December 1, 8 and 9, and the intention to set a date at next week's ACDC if Hoodi went well. It was a notice, not a decision, and the delay of the second testnet leaves it up in the air.

There is a reason the delay weighs more on mainnet than on Hoodi. The contracts that concern ChainSecurity are those already running in production, and a testnet only reproduces part of that tangle. The later the fork, the more time there is to locate the affected protocols and warn them, which is exactly what the ongoing disclosure is doing.

For layer 2s the postponement does not move the main needle. On October 4 the price of blobs rose by around 400%, and that cost depends on data space, not on execution gas. In Glamsterdam's list of 18 EIPs there is no proposal that expands the per-block blob quota, so the data bottleneck for layer 2s remains where it was. Aztec, which closes the vote on its jump to v6 on October 24, has already put Glamsterdam gas constants into the next version.

What the postponement buys and what it leaves open

Four days before the date, a testnet is the cheap place to discover that a pricing change breaks contracts. Repricing is the hardest thing to rehearse, because it depends on combinations of calls that no one writes in a unit test.

The next signal is in the call schedule and in the client notes, not in a date that does not exist today. Until Hoodi has an epoch, operators live with half-updated configurations and mainnet teams with a fork without a day. October 26 at 17:42:48 UTC is no longer an Ethereum date.

José Maldonado

Author

José Maldonado

Activist and blogger on technology, free software and blockchain.

The latest
From Huang to Botín, from Zara to Byron: corporate responsibility in global capitalism

Keep reading