File _patchinfo of Package patchinfo.13252
<patchinfo incident="13252"> <issue tracker="bnc" id="1108393">sbd.service not being added to pacemaker.service 'Required=' deffinition</issue> <issue tracker="bnc" id="1168771">Cluster rolling upgrade is broken and could result in unexpected huge corosync ringid that could not be handled by pacemaker and dlm</issue> <issue tracker="bnc" id="1174915"></issue> <issue tracker="bnc" id="1171372">Fix handling of fence-agents parameters in pacemaker</issue> <issue tracker="bnc" id="1148236">sbd_start error when losing one of three sbd devices. ref:_00D1igLOd._5001iIw1BZ:ref</issue> <issue tracker="bnc" id="1155290">Error performing operation: Update does not conform to the configured schema</issue> <issue tracker="bnc" id="1154881">Resource shows started in maintenance mode</issue> <issue tracker="bnc" id="1160410">pacemaker fails with -fno-common</issue> <issue tracker="bnc" id="963674">Wrong option in sbd manpage</issue> <issue tracker="bnc" id="1143064">sbd cannot run as RR if systemd's CPUAccounting is enabled</issue> <issue tracker="bnc" id="1140065">SBD hogs CPU in HA Crowbar ECP deployments</issue> <issue tracker="bnc" id="1150429">sbd should report "open_device: Opening device /dev/xxx failed" at the first place</issue> <issue tracker="jsc" id="ECO-1611"/> <issue tracker="jsc" id="SLE-12243"/> <packager>yan_gao</packager> <rating>moderate</rating> <category>recommended</category> <summary>Recommended update for pacemaker, sbd</summary> <description>This update for pacemaker, sbd fixes the following issues: Changes in pacemaker: - fencer: Avoid possible use-of-NULL when parsing metadata. (bsc#1171372) - libstonithd: Make the assert message from stonith__device_parameter_flags() more clear. (bsc#1171372) - libstonithd: Respect `pcmk_host_argument=none` on `validate`. (bsc#1171372) - libstonithd: Add `port` or `plug` parameter according to metadata on `validate` if no `pcmk_host_argument` specified. (bsc#1171372) - Pacemaker Explained: update the default value and the description of `pcmk_host_argument`. (bsc#1171372) - fencer: Add `port` or `plug` parameter according to metadata for RHCS-style fence-agents. (bsc#1171372) - libstonithd: Add function to check supported parameters according to the metadata of a fence agent. (bsc#1171372) - stonith_admin: --delay is an optional option for --fence/--unfence/--reboot commands (jsc#ECO-1611, jsc#SLE-12243) - controller: Avoid possible use-of-NULL on logging fencing message. (jsc#ECO-1611, jsc#SLE-12243) - cts: Add `plug` parameter for fence_dummy agent. (bsc#1171372) - libstonithd: `plug` parameter of RHCS-style fence-agents is shown as non-required in the metadata. (bsc#1171372) - libstonithd: `action` parameter of RHCS-style fence-agents is shown as non-required in the metadata. (bsc#1171372) - libstonithd: Functionize fudging metadata of RHCS-style agent to make specific parameter non-required. (bsc#1171372) - libcrmcluster: Use uint64_t type for corosync ringid (membership id) when updating node state. (bsc#1168771) - fencer: Update cpg_topology_delay test to also verify pcmk_delay_base is added. (jsc#ECO-1611, jsc#SLE-12243) - controller: Requested priority fencing delay defaults to 0. (jsc#ECO-1611, jsc#SLE-12243) - stonith_admin: Set --delay option defaults to 0. (jsc#ECO-1611, jsc#SLE-12243) - fencer: Any delays from pcmk_delay_base/max are added to requested fencing delay. (jsc#ECO-1611, jsc#SLE-12243) - scheduler: Do not differentiate the case where all the nodes have equal priority. (jsc#ECO-1611, jsc#SLE-12243) - scheduler: Set priority-fencing-delay defaults to 0 meaning disabled. (jsc#ECO-1611, jsc#SLE-12243) - Pacemaker Explained: Document priority-fencing-delay cluster option. (jsc#ECO-1611, jsc#SLE-12243) - fencer: Add cpg_topology_delay test to verify enforced fencing delay with fencing topology. (jsc#ECO-1611, jsc#SLE-12243) - fencer: Handle any enforced fencing delay. (jsc#ECO-1611, jsc#SLE-12243) - stonith_admin: Add --delay option to support enforced fencing delay. (jsc#ECO-1611, jsc#SLE-12243) - controller: Request fencing with any enforced priority fencing delay. (jsc#ECO-1611, jsc#SLE-12243) - libstonithd: Introduce fence_with_delay() operation. (jsc#ECO-1611, jsc#SLE-12243) - scheduler: Add regression test for priority-fencing-delay. (jsc#ECO-1611, jsc#SLE-12243) - scheduler: Implement priority-fencing-delay. (jsc#ECO-1611, jsc#SLE-12243) - scheduler: Add priority-fencing-delay cluster option. (jsc#ECO-1611, jsc#SLE-12243) - attrd: Properly declare global variables as extern in header. (bsc#1160410) - fencer: Do not require API registration for list and status commands. (bsc#1148236) - fencer: Improve error checking and log messages for API action requests. (bsc#1148236) - scheduler: Make sure cluster-wide maintenance-mode=true overrides per-resource settings (bsc#1154881) - cts-cli: Simplify and fix regexp to catch crm_time_as_string's output. (bsc#1155290) - cts-cli: Use extended regular expressions. (bsc#1155290) - cts-cli: Add tests for more crm_resource options. (bsc#1155290) - tools: Clear all prefer constraints when performing a move. (bsc#1155290) - tools: Fix moving a resource with a lifetime constraint. (bsc#1155290) Changes in sbd: - sbd-inquisitor: Refuse to start if any of the configured device names is invalid. (bsc#1174915) - scheduling: Complete overhaul. (bsc#1143064) - Doc: Add environment section to man-page. - agent: Correctly compare string values when calculating timeout. (bsc#1148236) - regressions.sh: Relaxed timeouts for tests under load. - tests: Add regression-tests using preload-library. - tests: Added preload-library for reboot interception. - sbd-inquisitor: Use crashdump timeout. - Build: switch back to serial test-harness. - Doc: mention crashdump message in usage note. - defaults: Make 15s timeout default for s390 consistently. - cmdline: Just use SBD_DEVICE if no devs from cmdline. - sbd.8.pod and -h help text add -vvv description. - sbd-md: Add a warning log if failed to open/read device on startup. (bsc#1150429) - agent: Log detailed errors for monitor failures. (bsc#1148236) - sbd-md: list/dump failures go to stderrL (bsc#1148236) - Avoid deprecated names for g_main-loop-functions. - sbd-pacemaker: Check for shutdown attribute on every cib-diff. - sbd-cluster: Periodically check corosync-daemon liveness. - sbd-pacemaker: Assume graceful exit if leftovers are unmanged. - sbd-common: Query rt-budget > 0 otherwise try moving to root-slice. (bsc#1143064) - systemd: Make pacemaker & dlm wait for sbd-start to complete. (bsc#1108393) - Fix node name parameter in manpage. (bsc#963674) </description> </patchinfo>
