On Tue, Oct 12, 2021 at 5:48 AM Regev Batchen <[email protected]> wrote:
> Hey Again, i'm still not able to handle this error that happens once a > day, all containers are up but api request to gqm gets 502 > I did get the full postgres query when this usually happened: > > postgres_gqm_1 | 2021-10-12T06:55:19.947186314Z 2021-10-12 06:55:19.938 > UTC [35] LOG: could not send data to client: Connection reset by peer > I don't think your issues are specifically related to Guacamole - I think there is something else driving network connectivity challenges between your containers, which is very odd since networking on a local system doesn't usually have much in the way of limits or instability. But, that's really what it looks like, here - between this error, where the peer is resetting the connection, and the previous gaucd one you posted about the error reading select, it really seems like a network problem. > postgres_gqm_1 | 2021-10-12T06:55:19.947212467Z 2021-10-12 06:55:19.938 > UTC [35] STATEMENT: SELECT > postgres_gqm_1 | 2021-10-12T06:55:19.947215854Z > guacamole_connection.connection_id, > postgres_gqm_1 | 2021-10-12T06:55:19.947218656Z > guacamole_connection.connection_name, > postgres_gqm_1 | 2021-10-12T06:55:19.947221509Z > parent_id, > postgres_gqm_1 | 2021-10-12T06:55:19.947224165Z > protocol, > postgres_gqm_1 | 2021-10-12T06:55:19.947226788Z > max_connections, > postgres_gqm_1 | 2021-10-12T06:55:19.947229387Z > max_connections_per_user, > postgres_gqm_1 | 2021-10-12T06:55:19.947232302Z > proxy_hostname, > postgres_gqm_1 | 2021-10-12T06:55:19.947234855Z > proxy_port, > postgres_gqm_1 | 2021-10-12T06:55:19.947237399Z > proxy_encryption_method, > postgres_gqm_1 | 2021-10-12T06:55:19.947239993Z > connection_weight, > postgres_gqm_1 | 2021-10-12T06:55:19.947242599Z > failover_only, > postgres_gqm_1 | 2021-10-12T06:55:19.947245130Z > MAX(start_date) AS last_active > postgres_gqm_1 | 2021-10-12T06:55:19.947247726Z FROM > guacamole_connection > postgres_gqm_1 | 2021-10-12T06:55:19.947250330Z LEFT JOIN > guacamole_connection_history ON guacamole_connection_history.connection_id > = guacamole_connection.connection_id > postgres_gqm_1 | 2021-10-12T06:55:19.947253528Z WHERE > guacamole_connection.connection_id IN > postgres_gqm_1 | 2021-10-12T06:55:19.947256208Z ( > postgres_gqm_1 | 2021-10-12T06:55:19.947258731Z > $1::integer > postgres_gqm_1 | 2021-10-12T06:55:19.947261284Z , > postgres_gqm_1 | 2021-10-12T06:55:19.947263751Z > $2::integer > postgres_gqm_1 | 2021-10-12T06:55:21.230273860Z , > postgres_gqm_1 | 2021-10-12T06:55:21.230276158Z > $25400::integer > postgres_gqm_1 | 2021-10-12T06:55:21.230278327Z ) > postgres_gqm_1 | 2021-10-12T06:55:21.230280533Z GROUP BY > guacamole_connection.connection_id > > Is this helping in any way? My next step is gonna be upgrading the kernel > and libs on the host and if that doesn't work to move from docker to > kubernetes :( > > I mean, you can try those things, but I would spend time figuring out why network traffic is getting dropped to the point of causing these issues. Do you see elevated error rates on any of the interfaces in your environment? Any messages - not necessarily Guacamole-related - that point to traffic getting dropped/stopped/blocked/lost? > ERRINFO_DISCONNECTED_BY_OTHER_CONNECTION (0x00000005):Another user connected to the server, forcing the disconnection of the current connection. This is definitely not Guacamole-related, and is exactly what the error message describes - some other user connected to the server and forced the session to log off. This is a RDP server message, and Guacamole is just forwarding the error information through. -Nick >
