[
https://issues.apache.org/jira/browse/HDFS-17285?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
liuguanghua updated HDFS-17285:
-------------------------------
Description:
When dfsrouter start, it enters safe mode. And it will cost 1min to leave.
The log is blow:
14:35:23,717 INFO
org.apache.hadoop.hdfs.server.federation.router.RouterSafemodeService: Leave
startup safe mode after 30000 ms
14:35:23,717 INFO
org.apache.hadoop.hdfs.server.federation.router.RouterSafemodeService: Enter
safe mode after 180000 ms without reaching the State Store
14:35:23,717 INFO
org.apache.hadoop.hdfs.server.federation.router.RouterSafemodeService: Entering
safe mode
14:35:24,996 INFO
org.apache.hadoop.hdfs.server.federation.router.RouterSafemodeService: Delaying
safemode exit for 28721 milliseconds...
14:36:25,037 INFO
org.apache.hadoop.hdfs.server.federation.router.RouterSafemodeService: Leaving
safe mode after 61319 milliseconds
It depends on these configs.
DFS_ROUTER_SAFEMODE_EXTENSION 30s
DFS_ROUTER_SAFEMODE_EXPIRATION 3min
DFS_ROUTER_CACHE_TIME_TO_LIVE_MS 1min (it is the period for check safe mode)
Because in safe mode dfsrouter will reject write requests, so it should be
shorter in check period if refreshCaches is done. And we should remove
DFS_ROUTER_CACHE_TIME_TO_LIVE_MS form RouterSafemodeService.
was:
When dfsrouter start, it enters safe mode. And it will cost 1min to leave.
The log is blow:
14:35:23,717 INFO
org.apache.hadoop.hdfs.server.federation.router.RouterSafemodeService: Leave
startup safe mode after 30000 ms
14:35:23,717 INFO
org.apache.hadoop.hdfs.server.federation.router.RouterSafemodeService: Enter
safe mode after 180000 ms without reaching the State Store
14:35:23,717 INFO
org.apache.hadoop.hdfs.server.federation.router.RouterSafemodeService: Entering
safe mode
14:35:24,996 INFO
org.apache.hadoop.hdfs.server.federation.router.RouterSafemodeService: Delaying
safemode exit for 28721 milliseconds...
14:36:25,037 INFO
org.apache.hadoop.hdfs.server.federation.router.RouterSafemodeService: Leaving
safe mode after 61319 milliseconds
It depends on these configs.
DFS_ROUTER_SAFEMODE_EXTENSION 30s
DFS_ROUTER_SAFEMODE_EXPIRATION 3min
DFS_ROUTER_CACHE_TIME_TO_LIVE_MS 1min (it is the period for check safe mode)
Because in safe mode dfsrouter will reject write requests, so it should be
shorter in check period if refreshCaches is done. And we should be separted
DFS_ROUTER_CACHE_TIME_TO_LIVE_MS form RouterSafemodeService.
> [RBF] Decrease dfsrouter safe mode check period.
> ------------------------------------------------
>
> Key: HDFS-17285
> URL: https://issues.apache.org/jira/browse/HDFS-17285
> Project: Hadoop HDFS
> Issue Type: Improvement
> Reporter: liuguanghua
> Priority: Minor
>
> When dfsrouter start, it enters safe mode. And it will cost 1min to leave.
> The log is blow:
> 14:35:23,717 INFO
> org.apache.hadoop.hdfs.server.federation.router.RouterSafemodeService: Leave
> startup safe mode after 30000 ms
> 14:35:23,717 INFO
> org.apache.hadoop.hdfs.server.federation.router.RouterSafemodeService: Enter
> safe mode after 180000 ms without reaching the State Store
> 14:35:23,717 INFO
> org.apache.hadoop.hdfs.server.federation.router.RouterSafemodeService:
> Entering safe mode
> 14:35:24,996 INFO
> org.apache.hadoop.hdfs.server.federation.router.RouterSafemodeService:
> Delaying safemode exit for 28721 milliseconds...
> 14:36:25,037 INFO
> org.apache.hadoop.hdfs.server.federation.router.RouterSafemodeService:
> Leaving safe mode after 61319 milliseconds
> It depends on these configs.
> DFS_ROUTER_SAFEMODE_EXTENSION 30s
> DFS_ROUTER_SAFEMODE_EXPIRATION 3min
> DFS_ROUTER_CACHE_TIME_TO_LIVE_MS 1min (it is the period for check safe mode)
> Because in safe mode dfsrouter will reject write requests, so it should be
> shorter in check period if refreshCaches is done. And we should remove
> DFS_ROUTER_CACHE_TIME_TO_LIVE_MS form RouterSafemodeService.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
