Details
-
Bug
-
Status: Resolved
-
Minor
-
Resolution: Fixed
-
2.5.0
-
java version "1.8.0_121"
Java(TM) SE Runtime Environment (build 1.8.0_121-tdc1-b13)
Java HotSpot(TM) 64-Bit Server VM (build 25.121-b13, mixed mode)AD Domain Controllers
LDAP v.3
2012 R2 OS
-
Patch
Description
Ambari server configured to use "secure" ldap authentication.
authentication.ldap.primaryUrl=********:636
authentication.ldap.useSSL=true
We call the ldap_sync_events REST endpoint frequently to synchronize existing groups and a specific list groups. We had no issues with this until mid-October at which point we began to see:
"status" : "ERROR",
"status_detail" : "Caught exception running LDAP sync. simple bind failed: **********:636; nested exception is javax.naming.CommunicationException: simple bind failed: **********:636 [Root exception is java.net.SocketException: Connection reset]",
Troubleshooting:
- We saw random success and failure when attempting to sync a single group.
- With useSSL=false and an updated port ldap sync was consistently successful.
Cause:
- By default, ldap connection only uses pooled connections when connecting to a directory server over LDAP. Enabling SSL causes it to disable the pooling, resulting in poorer performance and failures due to connection resets.
- Around mid-October we increased the number of groups defined on the system (50+), this pushed us outside the "safe zone".
Fix:
Enable the SSL connections pooling by adding the below argument to startup options.
-Dcom.sun.jndi.ldap.connect.pool.protocol='plain ssl'
Reference:
https://confluence.atlassian.com/jirakb/connecting-jira-to-active-directory-over-ldaps-fails-with-connection-reset-763004137.htm
https://docs.oracle.com/javase/jndi/tutorial/ldap/connect/config.html
