+++ This bug was initially created as a clone of Bug #1122923 +++ Description of problem: When kdump configuration is updated using UI, the whole file is overwritten, so custom changes, which are impossible to change using UI, are lost. This behavior breaks kdump integration feature in oVirt 3.5, which sets those custom options during host deploy. Version-Release number of selected component (if applicable): ovirt-node-iso-3.5.0.ovirt35.20140707.el6 How reproducible: 100% Steps to Reproduce: 1. Login as root, add custom changes to /etc/kdump.conf and restart kdump service so these changes are applied: default reboot ext4 /dev/HostVG/Data path /core fence_kdump_nodes mperina.brq.redhat.com fence_kdump_args -p 7410 -i 5 2. Login as admin and change settings kdump settings using UI (for example change kdump target to NFS) 3. Login as root and check content of /etc/kdump.conf Actual results: Content of /etc/kdump.conf after step 3.: default reboot nfs str-02.rhev.lab.eng.brq.redhat.com:/mnt/export/nfs/170/nfs04 Expected results: Content of /etc/kdump.conf after step 3. when bug fixed: default reboot nfs str-02.rhev.lab.eng.brq.redhat.com:/mnt/export/nfs/170/nfs04 fence_kdump_nodes mperina.brq.redhat.com fence_kdump_args -p 7410 -i 5 Additional info:
RHEV-H QE can reproduce this issue. Test version: ovirt-node-iso-3.5.0.ovirt35.20140707.el6.iso ovirt-node-3.1.0-0.0.master.20140707.git2f40d75.el6.noarch Test steps: 1. Login as root, add custom changes to /etc/kdump.conf and restart kdump service so these changes are applied: default reboot ext4 /dev/HostVG/Data path /core fence_kdump_nodes mperina.brq.redhat.com fence_kdump_args -p 7410 -i 5 2. Login as admin and change settings kdump settings using UI to nfs 3. Login as root and check content of /etc/kdump.conf The test result same as above.
Test version: rhev-hypervisor6-6.6-20141212.0.iso(RHEV-H 6.6 for RHEV3.5) ovirt-node-3.1.0-0.34.20141210git0c9c493.el6.noarch rhev-hypervisor7-7.0-20141212.0.iso(RHEV-H 7.0 for RHEV3.5) ovirt-node-3.1.0-0.34.20141210git0c9c493.el7.noarch Test steps: 1. Login as root, add custom changes to /etc/kdump.conf and restart kdump service so these changes are applied: default reboot ext4 /dev/HostVG/Data path /core fence_kdump_nodes mperina.brq.redhat.com fence_kdump_args -p 7410 -i 5 2. Login as admin and change settings kdump settings using UI to nfs 3. Login as root and check content of /etc/kdump.conf Test result: After step2, An error occurred with follow error: ┌───────────────────────────────────────────────────────── An error occurred ─────────────────────────────────────────────────────────┐ │An error occurred │ │Unable to save to file! │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │< Close > │ │ so need assigned this bug again
In order to not use a template, but reliably match and set values, we're using augeas, which is failing to parse the file (so it refuses to set or save values). Augeas does not support fence_kdump_nodes or fence_kdump_args. Ill file a bug against augeas about this and block against it. Are there any other arguments can we expect to see?
Options fence_kdump_nodes and fence_kdump_args were added into standard kexec-tools distribution in Fedora 20, RHEL 6.6 and RHEL 7.1 to support oVirt kdump integration. And only fence_kdump_* options are altered by oVirt host deploy (along with some comment lines).
Though the support landed earlier, support hasn't made it to augeas yet. I asked because I'm curious whether there are any other options which augeas doesn't currently support but may be used.
Moving to 3.5.1 because this issue doesn't happen in regular flows.
This has a complex fix and we will fix this for zstream only if a customer reports this issue. Please reopen if that is the case.