Last modified: 2014-06-20 16:54:03 UTC
[04:17] < Reedy> running sync-common on fenari gives [04:18] < Reedy> rsync: failed to connect to localhost (127.0.0.1): Connection refused (111) [04:18] < Reedy> rsync error: error in socket IO (code 10) at clientserver.c(122) [Receiver=3.0.9] [04:18] < Reedy> sync-common tin seemingly WFM The configuration in /srv/deployment/scap/scap/scap.cfg (yo dawg...) specifies localhost as the default master rsync server and then overrides this to tin.eqiad.wmnet for the TLD of wmnet. The lookup of the proper config value is driven by the fully qualified domain name of the host running the script. Some production hosts are externally accessible and have fully qualified domain names like `fenari.wikimedia.org` which only match the default settings. This leads to them attempting to sync files from themselves rather than tin (or another deployment host) when `sync-common` is called without providing a host list.
Change 140726 had a related patch set uploaded by BryanDavis: Add configuration for wikimedia.org TLD https://gerrit.wikimedia.org/r/140726
Change 140726 merged by jenkins-bot: Add configuration for wikimedia.org TLD https://gerrit.wikimedia.org/r/140726
Verified fix by running `/usr/local/bin/sync-common` on fenari without providing any explicit rsync source host. fenari:~ bd808$ /usr/local/bin/sync-common 16:48:49 Copying to fenari.wikimedia.org from tin.eqiad.wmnet 16:48:49 Started rsync common 16:52:31 Finished rsync common (duration: 03m 42s)
