Autopilot consumes 100% of CPU

[JaviLib]

Background

After enabling autopilot (and wtcclient, although I don't know if it is related) with the following options:

autopilot.active=1
autopilot.maxchannels=1000
autopilot.allocation=1
autopilot.minchansize=900000
autopilot.maxchansize=1010000
;Autopilot.OpenChanFeeRate=11
autopilot.conftarget=1000
autopilot.private=false
;autopilot.heuristic=preferential:1
autopilot.heuristic=top_centrality:1.0
wtclient.active=true

My CPU goes 100% most of the time.

Your environment

• version of lnd 0.13.1
• which operating system (uname -a on *Nix) Linux debian 5.10.0-8-amd64
  

[Roasbeef]

Are you able to extract a CPU profile? https://github.com/lightningnetwork/lnd/blob/a329c8061266418bccd797aa5d7d277c736ee930/docs/debugging_lnd.md#capturing-pprof-data-with-lnd

If you modify allocation to say, 0.90 does the issue disappear?

What do you generally see in the logs? Particularly if you make the ATPL sub-system to the trace log level?

[JaviLib]

Yes, after decreasing allocation to 0.90 the problem disappears.
Do you think it is still a bug? It should stop looking for channels if there is not enough onchain funds. It is absurd to continue to waste CPU if there is no more funds to open channels. In fact I think searching for channels for autopilot should be a low priority task, no need to waste so much CPU even if there are enough funds.

[bhandras]

Did you have available funds in your wallet above autopilot.minchansize?

[JaviLib]

Did you have available funds in your wallet above autopilot.minchansize?

I did in the beginning but after many open channels by autopilot it didn’t.