You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I have a three node etcd cluster, which is working fine in general. But if I execute etcdctl member list with multiple endpoints, the etcd nodes log in irregular intervals the following errors at warning level:
The error doesn't occur on every execution. If I use just two endpoints, the message still appears sometimes. If I use just one endpoint, the error won't show up. It doesn't matter which endpoint I use and also the order doesn't matter. The etcdctl command itself returns fine in every case.
The etcd cluster has no data and no other load. The only command which is executed on the cluster, is the etcdctl member list command. If I stop my tests, the log stays clean.
If I enable debug log, I get additional messages, which I guess are expected:
I know, there are similar issue about this topic. But I think, none of them covers this specific issue enough or got closed without fixing the problem. The tls setup itself seems to work perfectly fine, otherwise I think I would get deterministic error messages. All three nods themself are working without any problems.
I know about #9165, which is more related to logs from "caller":"grpclog/grpclog.go:51" and I think won't cover this problem.
My best guess is, that one node responds faster than the others and etcdctl closes the connection to the remaining ones. And if the timing is bad, the error is logged.
Environment:
OS: Debian Buster
etcd Version: 3.5.0-pre
Git SHA: eee8dec (master from yesterday)
Go Version: go1.15.3
Go OS/Arch: linux/amd64
How to reproduce:
I created three etcd nodes on different virtual machines. I used certificates created after this manual: etcd/hack/tls-setup/README.md
Bash loop from an external system, using the client certificate which was created for the first node. Loop was executed from another system than the etcd servers:
while true; do ./etcdctl --debug --endpoints https://10.88.105.183:2379,https://10.88.104.50:2379,https://10.88.105.173:2379 --cert infra0.pem --key infra0-key.pem --cacert ca.pem member list; sleep 1; done
Resulting Logs on the three etcd nodes from 8:38:00 to 8:40:00, with the above loop.
First Node:
The text was updated successfully, but these errors were encountered:
thechristschn
changed the title
Rejected connection, Error EOF with multiple endpoints
Rejected connection, Error EOF when using etcdctl with multiple endpoints
Oct 27, 2020
I think there is a larger issue at hand are you on 3.4.13?
I cant get my cluster to communicate at all. Not even with http. Localhost while the port is listening is not responding to any queries. etcdctl times out.
This issue has been automatically marked as stale because it has not had recent activity. It will be closed after 21 days if no further activity occurs. Thank you for your contributions.
I have a three node etcd cluster, which is working fine in general. But if I execute
etcdctl member list
with multiple endpoints, the etcd nodes log in irregular intervals the following errors at warning level:The error doesn't occur on every execution. If I use just two endpoints, the message still appears sometimes. If I use just one endpoint, the error won't show up. It doesn't matter which endpoint I use and also the order doesn't matter. The etcdctl command itself returns fine in every case.
The etcd cluster has no data and no other load. The only command which is executed on the cluster, is the
etcdctl member list
command. If I stop my tests, the log stays clean.If I enable debug log, I get additional messages, which I guess are expected:
I know, there are similar issue about this topic. But I think, none of them covers this specific issue enough or got closed without fixing the problem. The tls setup itself seems to work perfectly fine, otherwise I think I would get deterministic error messages. All three nods themself are working without any problems.
I know about #9165, which is more related to logs from
"caller":"grpclog/grpclog.go:51"
and I think won't cover this problem.My best guess is, that one node responds faster than the others and etcdctl closes the connection to the remaining ones. And if the timing is bad, the error is logged.
Environment:
OS: Debian Buster
etcd Version: 3.5.0-pre
Git SHA: eee8dec (master from yesterday)
Go Version: go1.15.3
Go OS/Arch: linux/amd64
How to reproduce:
I created three etcd nodes on different virtual machines. I used certificates created after this manual: etcd/hack/tls-setup/README.md
The startup command for the first node:
Bash loop from an external system, using the client certificate which was created for the first node. Loop was executed from another system than the etcd servers:
Resulting Logs on the three etcd nodes from 8:38:00 to 8:40:00, with the above loop.
First Node:
Second Node:
Third node:
The text was updated successfully, but these errors were encountered: