Closed linked0 closed 2 years ago
We call the getBlockHeaders
API on peers while catching up on blocks, which means that a NetworkClient calls the API on each connection that is in fact an RPCClient
instance. There happens a deadlock
in this line of RCPClient
while locking a connection on its connection poll.
I have been working on a RPC rework, which got rid of the connection pool. Lets see if this still happens after we merge that
OK, thanks.
I think that the deadlock happens always at the 4th trial on a connection, which is RPCConfig.concurrency
+ 1.
The PR could solve this issue. The blocks aren't generated in the TestNet now. Shouldn't we restart the Validators?
Solved by PR #3276
This error happens when I run a validator in the local machine or AWS instance.
The process of getting missing block sigs gets stalled in the second catch-up task after starting a validator from scratch, which means the first catch-up task succeeds and the second catch-up task never ends with the following log messages which are repetitive.