nv21 Network Upgrade, scope and timelines discussion

[luckyparadise]

[Updated - November 16, 2023]

Summarizing the various discussions on the timeline, scope and name of this upgrade for ease of reference.

Upgrade name:
Watermelon

Timeline:
Code freeze is 10 October
Calibnet is Oct 19
Mainnet is Dec 12

NOTE:
The implementer teams have decided to patch the Calibration network, fixing this issue where partitions with expired sectors needed to be compacted before they can be moved. To fix this issue, there needs to be a calib upgrade planned 8 November 2023 at 11:00 UTC.

Scope: The following FIPs are considered in-scope for this upgrade

• FIP0052: Increase Max Sector Commitment to 3.5 years
• FIP0059: Synthetic PoRep
• FIP0070: Allow SPs to move partitions between deadlines
• FIP0071: Deterministic State Access (IPLD Reachability)
• FIP0072: Improved event syscall API
• FIP0073: Remove beneficiary from the self_destruct syscall
• FIP0075: FVM Randomness FIP

[anorth]

FIP-0052 "Increase max sector commitment to 3.5 years" has been accepted for some time and was bumped from nv19. Some associated (non-FIP) fixes to accounting for sector lifetime will come with it as prerequisites.

[jennijuju]

There are a couple protocol bugs introduced by FIP-0045 implementation need to be fixed, and also the snap sector deal activation bug that was postponed by the nv19 upgrade.

overall, I would propose nv21 to be focused on hardening and clean up the network existing protocol.

[luckyparadise]

@jennijuju thank you! I see there could also be some other FIP-focused priorities based on the Notion table you are tracking. I believe you are planning to identify priorities from the slew of requests and perhaps share them in this thread.

[anorth]

Depending on the path chosen, I'd nominate filecoin-project/FIPs#719 for inclusion – data termination penalty, or reduction of mandatory deal collateral. It's most likely a small code change. The main reason is to get this relatively simple pre-req out of the way for more involved changes I expect to propose for the subsequent upgrade.

[kaitlin-beegle]

I like the idea of including this, though I am very sensitive about the viability of processing any FIPs for nv21 which affect the SP financial status quo.

Added a clarifying question, but generally, I would feel especially inclined to include a proposal like this if the proposal ends up specifying a reduction/elimination of the penalty.

[anorth]

I have now drafted a FIP for discussion #719 -> #726

[cryptonemo]

I believe FIP-0059 (Synthetic PoRep) was already discussed in the FIPs channel for inclusion in nv21. If it hasn't official been yet, it should be scoped for inclusion!

[luckyparadise]

Scope [Updated 15 September 2023]

---

FIPs that are being considered to be included in nv21

• FIP0052: Increase Max Sector Commitment to 3.5 years
• FIP0059: Synthetic PoRep
• FIP0070: Allow SPs to move partitions between deadlines
• FIP0071: Deterministic State Access (IPLD Reachability)
• FIP0072: Improved event syscall API
• FIP0073: Remove beneficiary from the self_destruct syscall
• FIP0075: FVM Randomness FIP
• FIPXXXX: Direct Data Onboarding
• FIPXXXX: Long Term market Cron Fix

[anorth]

The timeline for nv21 shifted quite a lot since this discussion started, with the decision to attempt only one further upgrade on 2023. I made these comments in the decision matrix, but neglected to propagate them here after that was resolved. Some changes that are actively being worked on and I believe should be in scope include:

• Actor events for the verified registry Actor events for the verified registry, miner, and market actors (FIP-0083) FIPs#754
• Direct data onboarding Direct data onboarding: Onboarding data without built-in market deals (FIP-0076) FIPs#730
• Long-term fix for market actor cron use, continuing Built-in market cron risk and mitigations (FIP-0060) FIPs#638
• Storage fees replace batch balancer, a new discussion forthcoming after Separate proof fees from execution gas and replace the batch balancer FIPs#557
• FIP-0067 PoRep security policy PoRep Security Policy & Replacement Sealing Enforcement (FIP-0067) FIPs#727

Most of these are in the category of needing partial or complete implementation before the FIP can be drafted, which is why they don't have open FIPs or numbers yet.

@Stebalien has also proposed a number of changes in FIP discussions for the FVM team, which I assume they are actively implementing.

[kaitlin-beegle]

@luckyparadise, re: https://filecoinproject.slack.com/archives/C01EU76LPCJ/p1691530482848819?thread_ts=1691353710.931949&cid=C01EU76LPCJ

[zhiqiangxu]

@luckyparadise Is the link for Storage fees replace batch balancer wrong?

[luckyparadise]

Direct data onboarding FIP and Long Term Market Cron Fix FIP are no longer in scope having not been able to meet Last Call date. They should be prioritized for nv22 early 2024.

[luckyparadise]

We have decided to descope FIP-0070 from the nv21 upgrade. This is in response to a critical bug that could impact network integrity. In prioritizing the network's stability, descoping FIP-0070 is the most prudent course of action at this juncture and has been agreed to by network implementers. Read more here

[luckyparadise]

Timelines

This is to provide updates and openly communicate the final dates for the nv21 Watermelon Network upgrade. I will also be summarizing all the points in the first box above.

Prior to agreeing on the dates for this upgrade, a decision matrix was developed to visually depict the various references, scope, and dependencies for an upgrade and when the upgrade should land. Option 3 was the preferred alternative. You can see the decision matrix here.

Engineering teams openly expressed interest to land mainnet Nov 7 to avoid Lab Week, Thanksgiving, and early travels in December.

Hence:

Code freeze - Oct 10
Calilb upgrade - Oct 19
Mainnet upgrade - Dec 12

[luckyparadise]

I am updating the timeline above to reflect a new change. All FIPs are to enter Last Call by 13 September 2023. All threads above have been updated.

[luckyparadise]

Based on FilDev events and engineering capacity constraints, there are changes to the timeline for nv21 Watermelon upgrade.
-Code Freeze: Oct 10, 2023

• Calibnet Upgrade: Oct 19, 2023
• Mainnet Upgrade: Nov 28, 2023
  These dates have been reflected across all tracking and planning threads.

[jennijuju]

please see an update from implementers here filecoin-project/community#74

[luckyparadise]

We are currently re-evaluating the deployment timeline for the nv21 upgrade. This decision, made after in-depth consideration, aims to ensure optimal network performance and security. This is at the request of network implementation teams. More information will be shared once finalized.
In addition, it has been agreed to descope FIP-0070 from the nv21 upgrade. This is in response to a critical bug that could impact network integrity. In prioritizing the network's stability, descoping FIP-0070 is the most prudent course of action at this juncture.

[luckyparadise]

Implementation teams have agreed to adjust the mainnet date upgrade date to 2023-12-12. Timelines have been updated above. For more information see here

[luckyparadise]

Network upgrade name

---

Watermelon

[kaitlin-beegle]

The upgrade will likely fall at the end of summer....what about watermelon??

It's sweet, it's juicy, and it's a universal favorite! 🍉🍉🍉

[luckyparadise]

I love watermelons! Watermelon upgrade!

[arajasek]

I'd like to propose just SUMMER as a name too! Double-down on the theme :D

[anorth]

It's winter for some (but watermelon is still tasty).

[jennijuju]

WATERMELON 🍉

[jennijuju]

Is anyone actively working on filecoin-project/FIPs#673?

[luckyparadise]

@AxCortesCubero is working on this no?

[AxCortesCubero]

Sorry, saw this now, I am not actively working on this. The last thing was a quick fix to the changes made by FIP-0057, but was stopped by @Stebalien but that's where I left it. I think the outcome was for this to be its own new project evaluating how several different FIPs have affected the Batch Balancer and if we need or can do any simple modifications to restore proper batch balancer functioning. We can discuss more about how much effort we should put in this, if we should take on this project, but not something we are working on at the moment.

[steven004]

Do we consider having a fix for the issue: filecoin-project/FIPs#689 , or filecoin-project/builtin-actors#1276 ?

We do need proveCommitAggregate when baseFee is high enough.

[jennijuju]

As you can see filecoin-project/builtin-actors#1276, the optimization part is being worked on. Our team is not prioritizing filecoin-project/builtin-actors#1277:

• The original issue is a protocol bug described in the issue and should eventually be fixed.
• However, one of the top complaints we hear from the SP community is that the onboarding cost is high - there are multiple efforts ongoing to see how we can reduce it from the protocol level.
• Now, while Remove deal/sector activation from cron builtin-actors#1277 is the right thing to do, given currently ProveCommit sector is activating deals for free, and it wont be possible to make deal activarion ProveCommitAggregation to be cheaper than free. In other words, today SP can choose to use ProveCommit over ProveCommitAggregation even when the base fee reaches the batch balancer, so to leverage the free activation for a much cheaper gas cost, fixing #1277 or Abnormal gas usage of `ProveCommitAggregate` compared with `ProveCommitSector` FIPs#689 today will remove that advantage.

With that being said, if the community prefers correctness soon than $$ saving "trick" - teams should reconsider priority. I sent a quick survey here.

Note that a proper fix for filecoin-project/builtin-actors#1277 will need a FIP. We also estimated 4 engineering weeks for the work. Given the governance process and lotus engineering bandwidth, it would NOT be delivered in the current proposed timeline if we were the team working on it. I would love to see Venus, Forest or anyone in the community to take a lead on this before us too!

[jennijuju]

(a quick fix would be for always using ProveCommitAggregation even if its a sector size 1 - that is temporary mitigation and will take less engineering time to deliver.

[jennijuju]

We do need proveCommitAggregate when baseFee is high enough.

At what basefee - provecommitaggreegation would be cheaper than doing provecommit in practical today?

[eshon]

Please don’t release the next Lotus upgrade during EthCC July 17-20, 2023. There will be a lot of Filecoin events around it.

[jennijuju]

I think Switching to new Drand mainnet network is something should be actively on the radar for Filecoin core devs - as iiuc there is a plan to shut down the existing drand network in 2024/2025 with the existing LOE. Filecoin implementation should implement and test thoroughly against the new Drand network on testnet before deploying it onto mainnet; the implementation and testing should happen sooner rather than later to avoid rushed actions upon Drand's deprecation.

[jennijuju]

Note that the details of this network switch is pending review and discussion, but I think it should be actively considered for nv21.

[Stebalien]

It also removes one of the blockers for faster block times.

[Fatman13]

Any core dev could please give some initial evaluation on how #774 could be implemented?

It would be much appreciated that we start early even that 774 hasn't been accepted yet as we learned from previous FIP iterations that preemptive involvement of protocol engineering is critical. It would be even better that if we could get the implementation started before the FIP is accepted like how it was done for SDM.

CC @Stebalien @anorth @jennijuju @arajasek

[anorth]

Yes, I will. At first glance I think the proposals are fairly straightforward at the code level. Their likely impact, however, is not straightforward to analyse.

[jennijuju]

add on to Anorth's comment; there is also not a concrete proposal yet.

To make it clear, SDM authors prototyped it out themselves in the beginning, and implementers only started coding on the FIP after a concrete FIP Draft was created with enough details (including motivation, concrete proposal, analysis and etc), with the ask/goal of catching design flaws before the FIP Draft moves to Last Call.

[jennijuju]

Initial evaluation of the idea is technically feasible and sounds reasonable.