[ https://issues.apache.org/jira/browse/HDFS-10244?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16111272#comment-16111272 ]
Daryn Sharp commented on HDFS-10244: ------------------------------------ #2 is not necessary. The are no "important IBRs" that warrant a DN excessively buffering IBRs, let alone assuming it knows the NN's heartbeat expiry. The DN can choose to discard IBRs at any time and simply send a FBR either directly, or via willful re-registration. The latter may be easier to accomplish, ex. trick the DN into thinking it hasn't registered. > Avoid accumulation of IBRs for SNN when the standby is down for more than > expected time > --------------------------------------------------------------------------------------- > > Key: HDFS-10244 > URL: https://issues.apache.org/jira/browse/HDFS-10244 > Project: Hadoop HDFS > Issue Type: Bug > Reporter: Brahma Reddy Battula > Assignee: Brahma Reddy Battula > > As per discussion in HDFS-9917 and > [comment|https://issues.apache.org/jira/browse/HDFS-9917?focusedCommentId=15213947&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-15213947] > from [~vinayrpet] > Avoid accumulation of IBRs when the standby is down for long time, can we > consider as below. > 1. IBRs for StandbyNN can have a threshold ( say 100K or 1Million IBRs ). > 2. Also not to loose any important IBRs, IBRs can be cleared when "the > threshold is reached AND 'lastIBR' is more than 'heartbeatExpiryInterval'. > i.e. DataNode is considered dead in Namenode side". In that case, for sure > re-Register() will be called on reconnection to running NameNode (if any). -- This message was sent by Atlassian JIRA (v6.4.14#64029) --------------------------------------------------------------------- To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org