02-20-2018 01:36 PM
02-20-2018 09:39 PM
You could write a python or shell script to backup your fsimage on a regular basis put it in cron tab to automate .
Yes you can take a copy of your fsimage without interrupting the service .
02-21-2018 11:52 AM
Thanks for your help, Please correcct me if my understanding is correct or not
When we say we can copy fsimage without interrupting the service via below command
hdfs dfsadmin -fetchImage backup_dir
The backup produced with above command will be consistent as while starting up name node,NameNode process reads the fsimage file and loads it to memory and also it applies if any edits present in journal nodes newer than the fsimage.In other case if jornal nodes are not available the its possible to lose data/changes occured in the interim.
Is there any best practices like in what frequency we can take metadata backup for prod hadoop clusters.
Could you please help me what will happenn in below scenario,
"When Standby Namenode is down for longer duaration , who does checkpointing operations", Is there any cases of loosing data or inconsistency if Active name node also crashes during this time?