Xud: Caching Channel Balance Thresholds

[kilrau]

We want to cache the following values:
<coin>_channel_min_in (default 15% of total channel capacity, all channels aggregated)
<coin>_channel_min_out (default 15% of total channel capacity, all channels aggregated)

This issue is just about caching the value, we'll take care of rebalancing & other actions like sending alerts in follow-up PRs. The % should be configurable either via xud.conf/cli paramter requiring a restart, better even with a separate cli call not requiring to restart xud. The cached values should be updated every 60s and immediately after each trade.

[kilrau]

Since both parties of a channel will reach the threshold at the same time (assuming both use the default), there needs to be a negotiation who takes over re-balancing/splicing.

[kilrau]

Could you help me think this through? @offerm @sangaman

[kilrau]

Assigned to you since you were dealing with similar checks in #675 @erkarl

Again, this issue is just to cache the value. Please give feedback so that we can move to implementation on this soon.

[sangaman]

Oh I think this would actually be closed by #850, can you confirm @erkarl ?

[erkarl]

@sangaman Yes, it feels pretty related.

@kilrau as of now we're caching the maximum outbound channel balance values (per currency) every 60 seconds. I'm not completely clear on the X% max in/out rule.

[sangaman]

Oh actually I think I overlooked that this is supposed to be dealing with rebalancing. That's a complicated issue as far as I know, could you provide more details of what you have in mind @kilrau ?

[kilrau]

Sry for the super-late reply.

caching the maximum outbound channel balance

Great - I see #850 via channelBalance is already aggregating outbound (local) balances of all channels for one currency! Now we'd need the same for the inbound (remote) balances, to do a sanity check there too - added a comment.

I'm not completely clear on the X% max in/out rule.

This issue is about defining thresholds to throw alerts (not in scope of this issue, I just opened #864 for this) when one side of the channel falls below a certain percentage of the channels total (inbound+outbound) capacity. But thanks for questioning this, all channels aggregated, how I originally wrote it doesn't make sense just yet. I believe we are going for a "direct channel" approach, hence we should throw alerts if one particular channels is unbalanced. These are the proposed cached values which should be updated regularly:
<coin>_channel_min_in (15% of total channel capacity)
<coin>_channel_min_out (15% of total channel capacity)

This is what this issue is about: caching (& updating) these 2 values. General "direct channels" discussion here.

Oh actually I think I overlooked that this is supposed to be dealing with rebalancing. That's a complicated issue as far as I know, could you provide more details of what you have in mind @kilrau ?

It's just about caching the channelBalance value, not dealing with any rebalancing. That's indeed complicated and for a later stage. For now we just want to alert the exchange.

[kilrau]

Does it make sense to you to cache these two values as preparation for #864 ? Or would you prefer to do it all at once - also fine! @erkarl

[erkarl]

Let's keep this issue for caching these values and deal with the alerts separately.

[kilrau]

Agree to skip inbound balance, because no direct lnd api call available, plus dependent on other nodes to open channels to my node anyways. So this issue is about implementing a configurable threshold on aggregated channel outbound balance. @erkarl

[rsercano]

Agree to skip inbound balance, because no direct lnd api call available, plus dependent on other nodes to open channels to my node anyways. So this issue is about implementing a configurable threshold on aggregated channel outbound balance. @erkarl

Hi @kilrau is this still valid, should we cache only outbound balance's configured percentage ?

[kilrau]

Hi @kilrau is this still valid, should we cache only outbound balance's configured percentage ?

LND 0.12 will return inbound balance with the channelbalance call. So we should do both. The PR is already merged, you can run lnd master to test.