Bug 1539113
| Summary: | crm_master does not work unless --lifetime is specified | ||
|---|---|---|---|
| Product: | Red Hat Enterprise Linux 7 | Reporter: | Ken Gaillot <kgaillot> |
| Component: | pacemaker | Assignee: | Ken Gaillot <kgaillot> |
| Status: | CLOSED ERRATA | QA Contact: | cluster-qe <cluster-qe> |
| Severity: | urgent | Docs Contact: | |
| Priority: | urgent | ||
| Version: | 7.4 | CC: | abeekhof, aherr, cfeist, cluster-maint, phagara |
| Target Milestone: | rc | Keywords: | Regression |
| Target Release: | 7.5 | ||
| Hardware: | All | ||
| OS: | All | ||
| Whiteboard: | |||
| Fixed In Version: | pacemaker-1.1.18-11.el7 | Doc Type: | Bug Fix |
| Doc Text: |
(This affects a small subset of customers, and thus does not need to be in the release notes.)
Cause: pacemaker-1.1.16-12.el7_4.3 introduced a regression in the crm_master tool such that it would not always pass an option to crm_attribute indicating a change in node attributes.
Consequence: Promotion of master resources would not occur when using custom resource agents that did not pass such an option (such as --lifetime) themselves when calling crm_master.
Fix: crm_master now always passes --lifetime to crm_attribute
Result: Promotion of master resources works correctly, even when used with custom resource agents that do not use the --lifetime option.
|
Story Points: | --- |
| Clone Of: | Environment: | ||
| Last Closed: | 2018-04-10 15:34:42 UTC | Type: | Bug |
| Regression: | --- | Mount Type: | --- |
| Documentation: | --- | CRM: | |
| Verified Versions: | Category: | --- | |
| oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |
| Cloudforms Team: | --- | Target Upstream Version: | |
| Embargoed: | |||
|
Description
Ken Gaillot
2018-01-26 17:18:13 UTC
Fixed upstream by commit a8b1ded before the fix (1.1.16-12.el7_4.7-94ff4df): * modified /usr/lib/ocf/resource.d/pacemaker/Stateful to NOT include the "-l reboot" option in crm_master call * created a master/slave "test" resource using the modified resource agent * all nodes remain slaves as the "master-test" nvpair is located under /cib/configuration/crm_config/cluster_property_set in the cib xml (whilst the cluster expects it to be stored under /cib/configuration/nodes/node/instance_attributes for master election purposes) implementing the workaround (ie. adding "-l reboot" to the ocf:pacemaker:Stateful RA again) results in the master/slave resource working as expected -- "master-test" nvpair gets stored either under /cib/configuration/nodes/node/instance_attributes (with "-l forever") or /cib/status/node_state/transient_attributes/instance_attributes (with "-l reboot") and one of the nodes being properly promoted to master according to its score value after the fix (1.1.18-11.el7-2b07d5c5a9): * using neither "-l reboot" nor "-l forever" in the master/slave RA results in crm_master (and subsequently crm_attribute) defaulting to setting a permanent node attribute * promotion to master works as expected * there is no negative side-effect of crm_master script always prepending "-l forever" to the options array passed to crm_attribute, as this will get overridden correctly if the RA passes custom lifetime option (cmdline args parsed in-order, later arg replaces earlier) * (as can be inferred from the previous point) the node score attribute is stored either as a permanent node attribute (by default & with "-l forever") or a transient one (if "-l reboot" is used) Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory, and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://access.redhat.com/errata/RHEA-2018:0860 |