Closed Ewemek closed 9 months ago
Well ! I've tried a few things, such as restoring --miner.notify and trying every version between v1.12.12 and v1.12.16. Ultimately I noticed that the problem did not happen when I waited a few minutes after stopping the node, before starting it again. The other notable change is that instead of having 0 peers, and struggling to get some (it took like 6+ hours to get only a few), now I have more than 50 in less than 1 minute after starting the node.
So now I feel that the problem is closely related to peering, maybe the rate limiting system needs tweaking ?
Thanks for the thoughtful report! I'll look at it right away and get back to you.
@Ewemek I've identified what I believe to be the problem and am drafting a patch.
Ultimately I noticed that the problem did not happen when I waited a few minutes after stopping the node, before starting it again.
This is the workaround I would have offered, and with it you shouldn't see any more spurious pauses. Only a minute's pause or so should do.
If you'd like, please email me at isaac@etccooperative.org and we can continue this conversation directly and in more depth.
Hi, I'm a pool operator and we've just updated from
v1.12.12
tov1.12.16
. We now face a problem where the node commits new sealing work very rarely. The pool is calling getWork() very often but still gets the same work, until a new one is committed. As you can see in the logs, the node did not generate a work for heights 18,717,898 to 18,717,890 (included) although we did ask for a new job in the meantime.We have enough peers (~170) and relevant options we use are:
--mine --miner.threads 0
. We recently dropped the--miner.notify
maybe this is related.But after investigation, I feel that our problem is related to this commit: https://github.com/etclabscore/core-geth/commit/6f51aed27141413b9a7b66145fa8e278fc3bff89. To be more specific, I have the impression that the mining process get stopped but never gets started back. In particular, the
shouldStart
is never actually used, I guess this is a regression but I can't know for sure it's related to my problem.System information
Geth version:
v1.12.16
Expected behaviour
New work should be committed at least when a block for a new height is processed.
Actual behaviour
Node is not committing new works when it should.
Steps to reproduce the behaviour
Run v1.12.16 with following options:
--mine --miner.threads 0
Backtrace