[ https://issues.apache.org/jira/browse/GUACAMOLE-919?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17072740#comment-17072740 ]
Douglas Heriot commented on GUACAMOLE-919: ------------------------------------------ [~mechanix] I think you're right - that's two separate issues. I think I've seen that JWK error when the time on the server was not configured correctly and the NTP service wasn't running. The separate thing with the SQL connection I think is showing it's connecting to pgBouncer. The error message "server conn crashed" is different to before and is what is showing up in my pgBouncer logs. The interesting option I'm setting in pgBouncer is PGBOUNCER_QUERY_WAIT_TIMEOUT=7 I'm pulling pgBouncer from docker hub pgbouncer/pgbouncer:1.12.0 [https://www.pgbouncer.org/config.html#dangerous-timeouts] I don't know what impact setting TCP keepalive has when it looks like the database is just silently ignoreing connections. For the past week we've been running a build of Guacamole set with socketTimeout and haven't had any issues! I haven't done a test whether it's pgBouncer or the socketTimeout that really fixed the issue - I think both options can acheive it. But testing reducing the socketTimeout time does make it quicker to start working than the pgBouncer alone option. I've just submitted a pull request with the change to pass this through: [https://github.com/apache/guacamole-client/pull/488] POSTGRES_SOCKET_TIMEOUT=5 > An I/O error occurred while sending to the backend > -------------------------------------------------- > > Key: GUACAMOLE-919 > URL: https://issues.apache.org/jira/browse/GUACAMOLE-919 > Project: Guacamole > Issue Type: Bug > Components: guacamole-auth-jdbc-postgresql > Affects Versions: 1.0.0 > Reporter: Mechanix > Assignee: Nick Couchman > Priority: Minor > Attachments: image-2020-01-27-15-19-26-634.png > > > Hi, > we use guacamole with postgresql and openid extension. Guacamole and guacd is > deployed inside a k8s cluster. > For some reason, the authentication doesn't succeed sporadically; there is > only a blank page and this error message in the guacamole log: > *[pool-1-thread-1] WARN o.a.i.d.pooled.PooledDataSource - Execution of ping > query 'SELECT 1' failed: An I/O error occurred while sending to the backend.* > I suspect there is a weird timeout happening between guacamole and postgresql > but couldn't figure out why. > Any hints are much appreciated. Thanks > > -- This message was sent by Atlassian Jira (v8.3.4#803005)