Ceph flag settings (cluster behavior settings)

1. Introduction

In many cases, it is necessary to set relevant behaviors for the cluster (for example, noout); without delving into the implementation principle, it is relatively easy to read this part, mainly in practice (ha)

2. Cluster behavior settings / cancel / view 2.1Cluster behavior settings # # you can only operate on one cluster behavior at a time. Multiple cluster behaviors need to be cancelled by using for loop # ceph osd set {behavior} 2.2. # # you can only operate on one cluster behavior at a time Multiple users need to use for loop # ceph osd unset {behavior} 2.3cluster behavior to view # # to view the overall status of the cluster, you can view the flag flag setting # ceph health 3. Cluster behavior details Flag description noin (tested) osd does not join crushnoout (tested) osd does not kick out crushnoup (tested) osd service is not automatically marked as available nodown (tested) osd service does not automatically mark stop (in some cases) The osd service is stopped and cannot be pulled automatically. It will take some time to pull it up manually, so in order not to migrate data, you can force all osd to provide services normally.) pause stops reading and writing services, but does not affect the osd self-test full cluster usage has reached its limit. Set this flag to force any data writing to be denied. But data readable norebalance (see English introduction) Ceph will prevent new rebalancing operations.nobackfill (tested) force organization of new data backfill norecover (tested) forcibly prohibit data migration noscrub (tested) No in-depth check-generally set to prevent performance impact nodeep-scrub (tested) No in-depth check-generally set Avoid affecting the performance of notieragent (not contacted yet, see the English introduction) Ceph will disable the process that is looking for cold/dirty objects to flush and evict