Do you have a secondary namenode running? Secondary NN is used for this purpose; Also, if you have HDFS HA enabled, this problem will also not occur.
Regards, *Stanley Shi,* On Tue, Jul 22, 2014 at 7:24 AM, Anfernee Xu <anfernee...@gmail.com> wrote: > Hi, > > For some reason, all PIDs file are missing in my cluster, I have to > manually kill all java processes on all machines, then I restarted the > HDFS, but it took so long time in applying changes in edit log file, so my > question is how can I reduce the delay? My understanding is as follows, > could someone please give some comments on it? > > hadoop dfsadmin -safemode enter > > #save current in-mem data to image file and reset edit log > hadoop dfsadmin -saveNamespace > > > -- > --Anfernee >