Hide Forgot
Description of problem: Installer etcd server.crt is created with wrong X509v3 Subject Alternative Name when host has 2 network interfaces. The X509v3 Subject Alternative Name will take the ip address of the 1st interface listed even when the ip variables is defined Version-Release number of selected component (if applicable): 3.0.x How reproducible: Run ose install to install a clustered ectd Define all variables for IP and etcd interface Make sure the host has 2 network interfaces, use the 2nd on listed Actual results: San Identifier show IP for 1st interface Expected results: Cert will be populated with right information as defined in /etc/ansible/hosts file Additional info:
Potential fix is to move to using hostnames https://github.com/openshift/openshift-ansible/pull/437 This was the way I was going with this but I think there's still some issues to sort out here https://bugzilla.redhat.com/show_bug.cgi?id=1267740
I've uncovered the issue. We were using the openshift.common.ip (from openshift_facts) instead of using the ip address derived from etcd_interface when creating the server certificates. I have a completely untested PR here: https://github.com/openshift/openshift-ansible/pull/702
Ryan, Can you test the above PR to see if it solves your issue?
QE, please test with the situation that Ryan mentions above. The PR here: https://github.com/openshift/openshift-ansible/pull/702 should resolve the issue he is seeing.
Hi Jason, QE could reproduce this issue with environment as mentioned by Ryan. Wrong IP address was picked up as the SAN identifier in server.crt, but when testing with non eth0 network interface, BZ#1267733 was involved. QE also tried with PR https://github.com/openshift/openshift-ansible/pull/702, ansible installer quit when creating the server csr for etcd cluster: TASK: [etcd_certificates | Create the server csr] ***************************** fatal: [10.66.79.107] => Failed to template IP:{{ ansible_devices['ansible_' ~ etcd_interface]ipv4.address }}: template error while templating string: expected token 'end of print statement', got 'ipv4' So I agree with Scott that using hostnames would finally resolve the two bugs. I'll set BZ#1267733 as blocker of this bug.
I've updated https://github.com/openshift/openshift-ansible/pull/702 to address the issue you hit, I must have missed pushing a fix to the variable reference for creating the SAN, please re-test.
Arrgh, I also referenced the variable on the local host instead of the variable on the host the certificate is being generated for, should be fixed now.
Test with https://github.com/detiber/openshift-ansible.git -b etcd_cert_fix Ansible installer aborted at: TASK: [etcd_certificates | Create the server csr] ***************************** ... 140190467962784:error:220A4076:X509 V3 routines:A2I_GENERAL_NAME:bad ip address:v3_alt.c:477:value={# item['ansible_' ~ item.etcd_interface].ipv4.address #} 140190467962784:error:22098080:X509 V3 routines:X509V3_EXT_nconf:error in extension:v3_conf.c:93:name=subjectAltName, value=IP:{# item['ansible_' ~ item.etcd_interface].ipv4.address #} So move this to Assigned.
I updated https://github.com/openshift/openshift-ansible/pull/702 to address some edge cases. We now test to make sure that we can read the variables we expect from the interface specified by etcd_interface for the specified host. If you are still seeing issues, please update with the inventory file you are using as well as the output of `ip addr` on the hosts.
Configure etcd_interface for the etcd hosts in the inventory file and test with https://github.com/detiber/openshift-ansible.git -b etcd_cert_fix Ansible installer failed at: TASK: [etcd_ca | command openssl req -config openssl.cnf -newkey rsa:4096 -keyout ca.key -new -out ca.crt -x509 -extensions etcd_v3_ca_self -batch -nodes -subj /CN=etcd-signer@{{ ansible_date_time.epoch }} -days 365 ] *** fatal: [10.66.79.125] => Failed to template {{}}: template error while templating string: Expected an expression, got 'end of print statement'
I updated the PR again with some sanity checking around the setting of etcd_interface. I also fixed the issue where it started to require etcd_interface be set. It looks like something changed (either in ansible or RHEL) that is now causing setting an ansible environment variable to an empty string with the command module was causing some tasks to fail under certain circumstances, so I just explicitly set the variable to a value now.
Test with https://github.com/openshift/openshift-ansible/pull/702 again, etcd ha cluster could be set up correctly with etcd_interface specified for the specified host. The IP address in SAN of server.crt is the address of specified etcd_interface. So move this to VERIFIED.