[
https://issues.apache.org/jira/browse/HBASE-29965?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18065241#comment-18065241
]
Kevin Geiszler commented on HBASE-29965:
----------------------------------------
[~andor], I noticed the config value is always in to
{{ConfigurationManager.notifyAllObservers()}} as {{{}false{}}}. This
{{notifyAllObservers()}} method is what calls {{onConfigurationChange()}} for
each observer. I tried on a local HBase deployment where I set both
{{hbase.global.readonly.enabled}} and {{hbase.block.data.cacheonread}} to
{{false}} on cluster startup. I then set them to {{true}} in the config, saved
the file, and ran {{update_all_config}} in the HBase shell.
{{hbase.block.data.cacheonread}} was updated to {{{}true{}}}, while
{{hbase.global.readonly.enabled}} stayed {{{}false{}}}.
> Unable to dynamically change readonly flag
> -------------------------------------------
>
> Key: HBASE-29965
> URL: https://issues.apache.org/jira/browse/HBASE-29965
> Project: HBase
> Issue Type: Sub-task
> Reporter: Andor Molnar
> Assignee: Kevin Geiszler
> Priority: Major
>
> # Modified global readonly -> true in hbase-site.xml
> # Executed {{update_all_config}} in hbase shell
> # See the following in the logs:
>
> {noformat}
> 2026-03-05T01:28:52,436 INFO
> [RpcServer.priority.RWQ.Fifo.read.handler=7,queue=1,port=16000]
> hbase.HBaseServerBase: Reloading the configuration from disk.
> 2026-03-05T01:28:52,436 INFO
> [RpcServer.priority.RWQ.Fifo.read.handler=7,queue=1,port=16000]
> conf.ConfigurationManager: Starting to notify all observers that config
> changed.
> 2026-03-05T01:28:52,450 INFO
> [RpcServer.priority.RWQ.Fifo.read.handler=7,queue=1,port=16000]
> balancer.BaseLoadBalancer: slop=0.2
> 2026-03-05T01:28:52,451 INFO
> [RpcServer.priority.RWQ.Fifo.read.handler=7,queue=1,port=16000]
> balancer.StochasticLoadBalancer: Loaded config; maxSteps=1000000,
> runMaxSteps=false, stepsPerRegion=800, maxRunningTime=30000, isByTable=false,
> CostFunctions=[RegionCountSkewCostFunction,
> PrimaryRegionCountSkewCostFunction, MoveCostFunction,
> ServerLocalityCostFunction, RackLocalityCostFunction, TableSkewCostFunction,
> StoreFileTableSkewCostFunction, RegionReplicaHostCostFunction,
> RegionReplicaRackCostFunction, ReadRequestCostFunction,
> CPRequestCostFunction, WriteRequestCostFunction, MemStoreSizeCostFunction,
> StoreFileCostFunction] , sum of multiplier of cost functions = 0.0 etc.
> 2026-03-05T01:28:52,451 INFO
> [RpcServer.priority.RWQ.Fifo.read.handler=7,queue=1,port=16000]
> normalizer.RegionNormalizerWorker: Normalizer rate limit set to
> unlimited{noformat}
> Looks like read-only mode doesn't get flipped for some reason
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
