hadoop-hdfs-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Satyam Singh <satyam.si...@ericsson.com>
Subject Re: One datanode is down then write/read starts failing
Date Mon, 28 Jul 2014 16:08:57 GMT
Yes, there is lot of space available at that instant.
I am not sure but i have read somewhere that we must have datanodes live 
 >= replication factor given at namenode at any point of time. If live 
datanodes get less than replication factor then this write/read failure 

In my case i have given replication factor 2 and initially i have 2 live 
Then i killed one of the datanode and number of live DN becomes 1 and 
still replication is 2.

Please comment.

On 07/28/2014 09:31 PM, Wellington Chevreuil wrote:
> Can you make sure you still have enough HDFS space once you kill this DN? If not, HDFS
will automatically enter safemode if it detects there's no hdfs space available. The error
message on the logs should have some hints on this.
> Cheers.
> On 28 Jul 2014, at 16:56, Satyam Singh <satyam.singh@ericsson.com> wrote:
>> Hello,
>> I have hadoop cluster setup of one namenode and two datanodes.
>> And i continuously write/read/delete through hdfs on namenode through hadoop client.
>> Then i kill one of the datanode, still one is working but writing on datanode is
getting failed for all write requests.
>> I want to overcome this scenario because at live traffic scenario any of datanode
might get down then how do we handle those cases.
>> Can anybody face this issue or i am doing something wrong in my setup.
>> Thanx in advance.
>> Warm Regards,
>> Satyam

View raw message