I want to bring up current struggles in Moonriver collation.
In the past months we’ve seen declining collator rewards and same trend in interest by larger operators.
As a reference StakeFish decided to just close down their operations for Moonriver and simply shut down their 4 Collator nodes. The team did not care much about the network and haven’t gracefully exited the collator set. That being said - even 30+ days after deciding to shut down their nodes, and after several pings in direct communication channels, they still occupy 4 slots in the collator set and by having all their nodes offline they slow down the network and make their stakers miss out on rewards.
For this reason I would propose to include an automated chilling of offline collators after a specific duration like 4 rounds. I proposed the same thing already like a year ago and would really love to see this implemented for the sake of performance and stability of the network.
It would be beyond awesome if we could squeeze this in for RT2600
p2p.org also announced their planned exit from Moonriver collation, which will mean that another 4 collator set slots getting vacant: x.com
Currently the waiting list for collators is hosting 8 collator candidates and I highly doubt that all of them are actually connected to an active and ready machine, as many of these spent quite a while on the waiting list.
So I would love to see some more entities on that list, but the raised minimum bond made this super rare to happen. I would love to propose to reduce the minimum bond again (in the past we said this should always be up for re-nogatiation and current state seems to make this mandatory) to 500 MOVR for more inclusion and a buffer on the waitinglist.
As this is just my opinion and a recommendation I’d ask everyone to weigh in with their opinions and thoughts.
Thanks for posting your concerns @dev0_sik as we at the Foundation have also been thinking about the situation.
I believe your suggestion for automatic chilling was considered but after some deliberation, there were some concerns that something completely automatic like that might inadvertently cause total network shutdown in some edge cases.
Instead, RT2600 actually includes a feature (enableMarkingOffline) that can be enabled/disabled with governance. Once enabled, it allows anyone to call parachainStaking.notifyInactiveCollator with an address of an inactive collator address. If the collator is truly inactive (there’s some validation there), then it is flagged as offline and effectively chilled. I think this is something that could be enabled on Moonriver along with RT2600 and solve that particular problem.
With regards to the minimum bond, I think lowering to 500 MOVR is a great idea. I believe all the current collator bonds are 500 MOVR or less anyway. We can always re-evaluate down the line if things change. I’d love to hear from other community members on this as well…
yeah, good point , sik. in fact, when an entity decides to sunset its services, it’s always beneficial if they simply submit the scheduleLeaveCandidates extrinsic. however, it seems that many cannot follow these straightforward actions. It becomes extremely necessary to utilize governance to move them to a chilling state, as a significant number of their delegators suffer from this issue. many of them don’t pay attention to the fact that their collator isn’t producing blocks, and they stake without receiving any rewards, which can be disappointing
It appears that the situation within the active set on Moonriver has sufficiently stabilized. currently, there are no issues with collators competing for a place in the active set. however, when market conditions change and entities with 4 collators shut down their nodes, the high level of self-bonding can pose an issue for new collators. as we previously discussed, we can adjust the self-bonding in accordance with changing market conditions and the situation in general
however, we should not greatly underestimate the self-bonding amount. I believe that the self-bonding of 1000 MOVR may be reasonable, since too low self-bonding can potentially lead to a repeat of the situation when collators compete for a spot in the active set, which, in turn, destabilizes the situation. but this is solely my point of view on this issue
Just to double-check a small detail:
The enableMarkingOffline is just done once through governance and if activated the parachainStaking.notifyInactiveCollator can be executed as a regular extrinsic, right?
I would absolutely LOVE to see this. Totally makes sense.
The Feb. 2023 increase to the minimum bond found in this link noted the bond “will be reviewed as deemed appropriate by the community”.
The total amount of staked MOVR (bond+stake - before the operators of these 8 nodes announced their intentions) needed to enter the active set had increased to ~20k – which is a much higher bond+stake than when the bond was increased in Feb 2023. Once the set rotates and new nodes join, it is expected that this total staked amount for the collator in 72nd place will rebalance to around 20k MOVR, which will provide a stabilizing effect to the active set. In the event future operators decide to move on, those positions should be more quickly backfilled with the bond reset to 500 MOVR. Therefore I support the move to decrease 500 MOVR.
Looking forward to the chilling effect RT2600 will have.
One other consideration is that the increase in bond was also an attempt to prevent a strategy that was seen whereby a node operator would buy their way into the set, and slowly decrease their self-stake until enough was reclaimed to buy in another node (via shifting delegations), and could effectively be done many, many times over time. A higher bond that needs to be locked in blunts this strategy. That said, I still support the decrease to 500 MOVR for bond.
I have mixed opinions on this topic. I understand the point of lowering the bond so that more collators are enticed to join the active set. Nevertheless, current market conditions also make it easier to enter with a higher bond.
Moreover, if you look into the current collator list, there are 5 collators that have a stake of 18k MOVR. This means that with the 10k MOVR bond, they need to get 8.8k MOVR just to be in the active set.
IMO - For the short term, we could think of reducing the collator set from 72 Collators to 64. This will boost by a tiny amount collator rewards among the remaining 64 collators, but still ensures the collator set is decentralized enough for the network.
IMO - This is the easiest short term solution for the problem, and then we can think of other more long term solutions to boost the collator set back to 72 after analyzing the impact of lowering the bond, etc.
Hey everyone–I have to side with Alberto’s suggestion on this one. Reducing the active set temporarily to 64 until we have a firm actionable plan makes the most sense to me. I strongly believe we should clearly state that the reduction is temporarily and provide a timeline so everyone can expect when the active set will be expanded again and have ample time to ensure their nodes are fully functional. I do want to stress that this suggestion shouldn’t be considered as an attempt to become more “centralized”, that is never the intent.
I am also in favor of the min bond reduction but personally don’t agree dropping to 500 MOVR. Back in Feb 2023 it was increased to align with the market price that was required when MOVR first went active. However I think the time has come that we no longer need to peg the min bond to the Day1 launch price anymore. I was thinking 2,000 MOVR, and no I don’t have any data to justify that amount. Its what I feel is a low entry barrier to encourage new user adoption but not too low to create a disaster in the active set.
A couple of things to note on this. At least in the linked discussion below, the view at that time from @artkaseman was that the diff between active collators between Moonriver and Moonbase shouldn’t be more than 8. If the number was lowered on Moonriver below 68 it would violate that previous view, but maybe it would be considered acceptable now.
Without speculating on price, looking at the economics of collating on Moonriver is an important aspect of this.
Currently a collator will expect to earn around 100 MOVR / month in rewards, or somewhere between $400-$500. This just about pays for professional grade bare metal server + 1 backup in a top tier data center, and says nothing for labor. We don’t know the reasons why P2P and StakeFish are ending their services, but I think it’s reasonable to assume the economics was a factor.
The point I’m trying to make is that if we don’t speculate on price, the economics of running a collator today would likely only be attractive new collators that are hobbyists with the technical skills to run a collator, who also happen to have 10k in bond available.
I don’t think the self bond is so important from the reason you’ve mentioned (quantity) since the amount to become active is much higher
Whether its 500 or 10000 MOVR self bond, you still need to raise about 20k MOVR in total to become active
However it is important from the complexity perspective by forcing new collators to lock 10k as self bond. This is a significant entrance barrier due to the fact that those 10k should be always locked
I mean if someone buys 10k MOVR and manage to raise another 10k somehow, and let’s say he is also a tech savvy, I’m not sure he’ll decide to launch a collator when he has to permanently lock his 10k as self bond (better to stake it, be liquid and without operational node hassles). Honestly, I don’t think if anyone of the current collators (except foundation of course) hold more than 10k MOVR (or similar amount), so expecting new collator not just to buy 10k MOVR but permanently lock it is too much I believe
Therefore, I recommend to reduce it back to a lower number (maybe not 500, but around it)
Everyone has made some really good points here but I’m not sure where this leaves us. I would say the window is rapidly closing if we want to get a change to the min bond in RT2600.
It sounds like everyone agrees that 10K MOVR is a lot of capital to put up relative to the income it generates and so it would make sense to lower it to something. Perhaps we can agree on something but higher than 500?
I believe we can change the active set size without a runtime upgrade so we can always do that later after the run time upgrade.
My (personal) thinking is that we lower the bond and observe. If we don’t get takers we can chill any problematic collators and then figure out if we should reduce the active set.
As @Jim_CertHum pointed out, I don’t think we’d want to shrink the active set too much as Moonriver serves to alert us to any potential scale challenges.
A quick poll I think would be the best way forward to judge the option with the most support.
I propose that the the option with the most votes gets moved forward, and if an alternative option (via comment) gets enough support, that one is implemented. I’ve tried to capture all of the views expressed so far, but I think a bias towards action is needed here, too, and so limited the poll to close in 24hrs.
There are 72 active collator slots on Moonriver, and in the near term, 8 of those will be taken by inactive collators (it’s unclear is P2P will manually remove themselves as a candidate). This has a negative impact on block production. What action (if any) should the community take?
Do nothing - leave minimum bond at 10k, with RT2600 Stakefish nodes can be removed, as well as P2P if they don’t act.
Proactively reduce minimum bond to 500 MOVR, opening up the possibility for more collators to become eligible.
Proactively reduce minimum bond to 1000 MOVR, opening up the possibility for more collators to become eligible.
Decrease the active set by <8> active nodes, leaving everything else the same.
Decrease the active set by active nodes, leaving everything else the same (I’ll comment on what should be)
Separately, we still have the issue of stakefish collators that are offline. We could wait for RT2600 on Moonriver and then do a governance action to enable the “enableMarkingOffline” feature.
However, the updated target for RT2600 on Moonriver is November 28th and only then can a referendum be made to enable the feature so we’re looking at the better part of a month where the stakefish collators impact block production.
