[ https://issues.apache.org/jira/browse/HDFS-8789?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16911798#comment-16911798 ]
Wei-Chiu Chuang edited comment on HDFS-8789 at 8/20/19 11:06 PM: ----------------------------------------------------------------- Very quickly went through the patch. (1) if this tool runs for an extended period of time, say more than a day, it may fail in a Kerberized environment since it does not renew kerberos credentials. It doesn't renew block keys. Just like HDFS-11741. (2) I'm still not clear about the process of running this tool -- is an administrator supposed to run this tool before restarting NN and enabling BlockPlacementPolicyWithUpgradeDomain? But NameNode would treat the new replica as mis-replicated block replica, and recalculate, wouldn't it? The migrator has to finish migration faster than NameNode can move replicas back. (3) What if I simply skip the migrator step, and enable UD. NameNode would see a huge number of mis-replicated blocks, but it should eventually move the replicas to the correct location. Of course, we need to fix HDFS-14637 first. (4) i wonder if it makes sense to integrate migrator to balancer/mover. There are already balancer/mover/diskbalancer. Having another similar tool in the arsenal can be a support burden. From what I see, balancer doesn't support anything other than the default placement policy (BlockPlacementPolicyDefault). Is the migrator tool sort of like a balancer designed to support UD? was (Author: jojochuang): Very quickly went through the patch. (1) if this tool runs for an extended period of time, say more than a day, it may fail in a Kerberized environment since it does not renew kerberos credentials. Just like HDFS-11741. (2) I'm still not clear about the process of running this tool -- is an administrator supposed to run this tool before restarting NN and enabling BlockPlacementPolicyWithUpgradeDomain? But NameNode would treat the new replica as mis-replicated block replica, and recalculate, wouldn't it? The migrator has to finish migration faster than NameNode can move replicas back. (3) What if I simply skip the migrator step, and enable UD. NameNode would see a huge number of mis-replicated blocks, but it should eventually move the replicas to the correct location. Of course, we need to fix HDFS-14637 first. (4) i wonder if it makes sense to integrate migrator to balancer/mover. There are already balancer/mover/diskbalancer. Having another similar tool in the arsenal can be a support burden. From what I see, balancer doesn't support anything other than the default placement policy (BlockPlacementPolicyDefault). Is the migrator tool sort of like a balancer designed to support UD? > Block Placement policy migrator > ------------------------------- > > Key: HDFS-8789 > URL: https://issues.apache.org/jira/browse/HDFS-8789 > Project: Hadoop HDFS > Issue Type: New Feature > Reporter: Chris Trezzo > Assignee: Chris Trezzo > Priority: Major > Attachments: HDFS-8789-trunk-STRAWMAN-v1.patch > > > As we start to add new block placement policies to HDFS, it will be necessary > to have a robust tool that can migrate HDFS blocks between placement > policies. This jira is for the design and implementation of that tool. -- This message was sent by Atlassian Jira (v8.3.2#803003) --------------------------------------------------------------------- To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org