Lotus also kills my router

[Beanow]

Consumer routers crashing / resetting / locking has been an issue that's plagued IPFS and friends for a long time.
See ipfs/kubo#3320

I've not seen it mentioned searching for this on lotus specifically, but wanted to point out this is the case for me. Regardless of what I set the available throttling to.

I tried several settings, even something as cramped as this:

[Libp2p]
  ConnMgrLow = 5
  ConnMgrHigh = 20
  ConnMgrGrace = "60s"

And yet every couple of minutes it will kill my router.
(My money is on it being due to hysterical dialing, not # of established connections)

Did I miss an option to further throttle lotus? Or did this issue carry over from the underlying tech?

[s0nik42]

What's your access bandwith ?

Monitoring the lotus node throughput, mine needs between 2-4MiB/s of bandwith. I never being able to keep it sync over my DSL backup access.

[Beanow]

500:40 Mbitps is the ISP contract I have, if that's what you mean.
Though note it's not that I'm lagging behind or saturating bandwidth. The modem actually crashes and resets, fully disrupting LAN, WiFi, etc.

[stuberman]

There are two primary P2P processes - one for blockrewards (.lotus) and one for deals (.lotusminer) - my connections are usually in the thousands. It is true that low end consumer routers will not tolerate this network activity. I had to upgrade my ISP's router from the default to a newer, more robust, model ARRIS model BGW320.

[Angelo-gh3990]

I agree with Stuberman, the amount of sessions is probably the issue here.

[s0nik42]

Average number of sessions is 400 for miner and 400 for node

[Beanow]

Well 400 would certainly trip the modem/router. If I use other protocols, like 400 torrent connections that would kill it as well.
However that's what the connection manager settings are supposed to be for right? (a capped 200 torrent connections will happily run 24/7)

Note I'm not trying to create a miner or gateway that has high connectivity or throughput needs. I'm just trying to use it as a full node (since it's currently the only option I know of with Ledger hw wallet support without requiring Google Chrome spyware).

[LexLuthr]

@Beanow I think at this point we have established that even if there is no activity from your side, the router still crashes because of the normal traffic in full node as well as lite-node. I think only option left is to setup the node on a more robust platform without a consumer grade router in between.

[Beanow]

@Beanow I think at this point we have established that even if there is no activity from your side, the router still crashes because of the normal traffic in full node as well as lite-node. I think only option left is to setup the node on a more robust platform without a consumer grade router in between.

If I wanted a short term fix, yes :] But I would encourage the team to see if this can be managed (/limited) better.
Whether it's individuals behind low spec routers, or 20 devs hammering an office network looking to replace S3, I would argue the punishing network requirements are problematic.

Specifically as I mentioned:

(My money is on it being due to hysterical dialing, not # of established connections)

Extremely aggressive dialing hundreds of peers "shouldn't" be necessary to make a storage deal for a 10KB cat picture for the tutorial right?

That said @LexLuthr, just want to mention I appreciate your help triaging!

[LexLuthr]

@Beanow I will check with our devs and see if there is anything that can we done. I will keep this discussion open for now and will get back to you with their answer.

[LexLuthr]

@Beanow I have asked internally but it seems there is no other solution but move to a more robust router.

[Beanow]

Bummer, well feel free to reach out if things improve later on. For now I will take that as not being quite ready for prime-time. With some luck my area will get consumer fiber next summer, perhaps that'll net me a more robust router, or the option to replace it as well.