ZmnSCPxj / clboss

Automated Core Lightning Node Manager
MIT License
211 stars 32 forks source link

"Can't recover from plugin failure, terminating." #151

Closed AndiMacTen closed 1 year ago

AndiMacTen commented 2 years ago

Hi, I moved my CLN node to another computer. Therefore I copied the .lightning folder to the new installation and the node came up successfully again. After installing clboss, it seems clboss is preventing the cln daemon to start again. Sadly there is not much logs I could share.

From the cli $ lightningd --daemon Can't recover from plugin failure, terminating.

Debug (log-level=debug log-level=debug:plugin-clboss)

DEBUG lightningd: Opened log file /home/username/.lightning/debug.log DEBUG plugin-manager: started(9086) /usr/local/libexec/c-lightning/plugins/autoclean DEBUG plugin-manager: started(9087) /usr/local/libexec/c-lightning/plugins/chanbackup DEBUG plugin-manager: started(9088) /usr/local/libexec/c-lightning/plugins/bcli DEBUG plugin-manager: started(9089) /usr/local/libexec/c-lightning/plugins/commando DEBUG plugin-manager: started(9090) /usr/local/libexec/c-lightning/plugins/fetchinvoice DEBUG plugin-manager: started(9091) /usr/local/libexec/c-lightning/plugins/funder DEBUG plugin-manager: started(9092) /usr/local/libexec/c-lightning/plugins/topology DEBUG plugin-manager: started(9093) /usr/local/libexec/c-lightning/plugins/keysend DEBUG plugin-manager: started(9094) /usr/local/libexec/c-lightning/plugins/offers DEBUG plugin-manager: started(9095) /usr/local/libexec/c-lightning/plugins/pay DEBUG plugin-manager: started(9096) /usr/local/libexec/c-lightning/plugins/txprepare DEBUG plugin-manager: started(9097) /usr/local/libexec/c-lightning/plugins/spenderp DEBUG plugin-manager: started(9098) /usr/local/libexec/c-lightning/plugins/cln-grpc DEBUG plugin-manager: started(9100) /usr/local/libexec/c-lightning/plugins/bookkeeper DEBUG plugin-manager: started(9101) /usr/local/bin/clboss UNUSUAL plugin-clboss: Killing plugin: timed out before replying to getmanifest DEBUG lightningd: io_break: check_plugins_manifests

I renamed the data.clboss and keys.clboss files but with no effect.

Only difference, old machine was cln release v0.12.0, the new installation is on cln current master (v0.12.0-106-g37c07dd).

Any idea what this issue could caused by and how ti get clboss running again? Thank you!

tsjk commented 2 years ago

I think I stumbled upon this as well. See https://gitlab.com/bitcoin/gentoo/-/issues/11, which I just wrote.

tsjk commented 2 years ago

PLUGIN_MANIFEST_TIMEOUT is defined in lightningd/plugin.c . Have you tried increasing that?

AndiMacTen commented 2 years ago

thanks for bringing this up. I increased that value by 10x, unfortunately with the same result as before.

tsjk commented 2 years ago

Was just a guess. But, that would have been too easy. I for one downgraded. Hopefully someone with some better knowledge of the code bases can give us a hint before we need to delve into it ourselves.

tsjk commented 2 years ago

I do wonder... What if you don't load the plugin at start, but load it after core-lightning has booted and is operational? I'll try that when I have time.

tsjk commented 2 years ago

I do wonder... What if you don't load the plugin at start, but load it after core-lightning has booted and is operational? I'll try that when I have time.

That didn't work either.

grubles commented 2 years ago

Running into the same issue when starting CLN (at commit 62bfed9a8df8731be44ba4e86afb08a5d28a4442):

UNUSUAL plugin-clboss: Killing plugin: timed out before replying to getmanifest
jonasnick commented 1 year ago

Same here (CLN 22.11rc2)

cdecker commented 1 year ago

The problem has been addressed on the CLN side for now by making the switch to non-numeric JSON-RPC calls go through the deprecation cycle: https://github.com/ElementsProject/lightning/pull/5727

Notice that this will temporarily allow clboss to use the numeric IDs, until they get removed in about 6 months time, so not a permanent fix, but should bridge the gap until clboss adds support for non-numeric IDs.

tsjk commented 1 year ago

I think this can be closed now.

cdecker commented 1 year ago

I don't think so: CLN added the nonnumeric ids as a temporary fix, deprecating numeric-only IDs. This means that after the deprecation cycle in ~6months we'll be facing the same issue again, unless clboss adds support for non-numeric IDs.

You can test support by starting lightningd with --allow-deprecated-apis=false and you'll see what lightningd expects after enforcing current deprecations

tsjk commented 1 year ago

Yes ok. It needs a new release. But the latest head is fixed. I think.

AndiMacTen commented 1 year ago

closed