Closed kallerosenbaum closed 6 years ago
~That's really strange behavior. It could be an invalid ~gossip_ping
message:
~But it's missing the ~id
which makes it invalid. And even if it were valid, it should not be forwarded to channeld
...
Edit: mixed up the message type, this is actually msgtype
3016, gossip_send_gossip
. See below.
So here's a tcpdump taken as sudo tcpdump -n -i eth0 -w dump.dat 'host c-5eea209f-74736162.cust.telenor.se and tcp port 9735'
, maybe it can be helpful? It covers a about 4 seconds worth of traffic.
Maybe it's better to turn on debug log level in lightningd? Tips on debugging arguments to lightningd?
Here's last 1000 lines of debug log (not taken simultaneously as the tcpdump above):
I'm afraid the transport layer is encrypted and we won't be able to decrypt it.
I'm actually pretty sure this issue is on our side, not Eclair. That message should not be sent there imho
@kallerosenbaum does this error persist across restarts? If yes, can you make a backup of the .lightning/gossip_store
file, remove it and then restart the node?
My error, it seems that this is msgtype=3016
, a.k.a., gossip_send_gossip
, attempting to send an empty gossip message.
Can you checkout and compile the issue-1587 branch which includes an assert that'll tell us how we got there?
It persists across restarts. I did a series of tests, in each test except 1, I used the backed up gossip_store
. :
gossip_store
: Problem dissappearsgossip_store
: Problem reappearsgossip_store
again, just to make sure: Problem does not appear (logfile issue-1587-branch.log)master
: Problem does not appeargit checkout 9d721ecb994813a84c9c631deb23117678bbd0d9
: Problem reappearsThe problem seems to be fixed somewhere after 9d721ecb994813a84c9c631deb23117678bbd0d9.
I'm uploading the issue-1587-branch eventhough it didn't happen there: issue-1587-branch.log.gz
Excellent work @kallerosenbaum. Still a bit curious about which commit accidentally fixed this issue. Could you send me a copy of your gossip_store
? It stores only publicly available information, so that should not be a concern. With it we can bisect and pin down the exact commit.
Sure: gossip_store_backup.gz
Thanks, I'll try to figure out what went wrong and report back :+1:
Hm, I still can't reproduce this, even with commit 9d721ecb994813a84c9c631deb23117678bbd0d9. But if I understood correctly master
doesn't have that issue anymore, right?
@rustyrussell pointed out that this was fixed in 518f8fc8d0231da6fe8396180b6d4be1a87ab746, cheers :wink:
Ok, thanks for sorting it out.
Issue and Steps to Reproduce
I have a channel between my eclair wallet and my c-lightning node. The connection seems to break every second and eclair reconnects. The channel status in Eclair toggles between OFFLINE and NORMAL and a the c-lightning log (of a newly started node) says:
Not sure if this is Eclair's, c-lightning's or my fault, but I report it on c-lightning because that's where I have some logging info.
I have some other channels open in c-lightning, and they all seem to work as expected. So I suspect it's an Eclair-clightning issue.
getinfo
output