大网数据广告活动
服务器运维问答

Pacemaker failure-timeout don't reset failcount

· 52CE 我爱测

I'm using Pacemaker 1.1.13 and Corosync 2.3.4 on Centos7. I've a problem with Master/Slave resource. There is meta attrs for my resource: migration-threshold=1 failure-timeout=10s but when the resource goes down, there is only on…

问题原文

原帖链接 · scooty · CC BY-SA 3.0

I'm using Pacemaker 1.1.13 and Corosync 2.3.4 on Centos7.

I've a problem with Master/Slave resource. There is meta attrs for my resource:

migration-threshold=1

failure-timeout=10s

but when the resource goes down, there is only one attempt to start it. Documentation says that attribute failure-timeout=10s should reset failcount every 10 seconds, but that does not happen, so resource never start.

Do You know anything about this problem? Maybe I'm doing something wrong? I'm sending my 'pcs status' below:

Cluster Name: webcluster
Corosync Nodes:
 10.121.100.101 10.121.100.102
Pacemaker Nodes:
 pm-node1 pm-node2

Resources:
 Master: Services-master
  Meta Attrs: failure-timeout=10s
  Group: Services
   Meta Attrs: migration-threshold=1
   Resource: Test (class=ocf provider=scooty type=test)
    Operations: start interval=0s timeout=20 (Test-start-interval-0s)
                stop interval=0s timeout=20 (Test-stop-interval-0s)
                monitor interval=10 role=Master timeout=20 (Test-monitor-interval-10)
                monitor interval=11 role=Slave timeout=20 (Test-monitor-interval-11)

Stonith Devices:
Fencing Levels:

Location Constraints:
Ordering Constraints:
Colocation Constraints:

Resources Defaults:
 migration-threshold: 1
 failure-timeout: 10
Operations Defaults:
 No defaults set

Cluster Properties:
 cluster-infrastructure: corosync
 cluster-name: webcluster
 dc-version: 1.1.13-10.el7_2.4-44eb2dd
 have-watchdog: false
 last-lrm-refresh: 1475145002
 no-quorum-policy: ignore
 start-failure-is-fatal: false
 stonith-enabled: false

已采纳解答

原帖链接 · Matt Kereczman · CC BY-SA 3.0

Depending on the type of failure, failure-timeout might not be enough to clean it up. Start and Stop operation failures are considered "fatal" and will not be automatically cleaned up by failure-timeout.

If you're having issues with a start operation failing, you can set the cluster property start-failure-is-fatal=false. Fencing/STONITH devices are the only way to recover from a stop failure.

Hope that helps.

摘自公开归档,保留原文语言,版式经整理,外部媒体与邮箱地址已移除。

资料来源与适用范围

资料来源
Server Fault
适用范围
以原帖系统、软件版本与发布日期为准(2016-09-29)
内容语言
中文;命令、路径和错误标识保留技术原文
本地收录日期
2026-11-15
授权说明
CC BY-SA 3.0

正文与所需资源由本站本地存储。查看来源及更新记录 ↗

请按适用版本和实际日志确认原因;本站没有对每一种系统或设备逐项执行文中的操作。