Support Questions

Find answers, ask questions, and share your expertise
Announcements
Celebrating as our community reaches 100,000 members! Thank you!

Slow Response From Standy NameNode when primary namenode is down,Response From Standby-NameNode is slow when active namenode is down

avatar
New Contributor

HI all, I have configured cluster server with Active-Standby NameNode. But whenever I turn of my active name node my response time is 1m 35 sec in average is this normal or am I missing some configuration. I am not using yarn only hdfs. Thankyou

,

In Cluster environment, whenever I turnoff my active-namenode the response from standby-namenode is slow on average of 1m 35 sec. Is this response time fine or it should be less. I haven't used yarn and mapreduce. Hadoop is only used for HDFS with kerberos authentication

1 ACCEPTED SOLUTION

avatar
@NLAMA

If you shutdown the Active Namenode, the Standby NameNode needs to become the Active NameNode, before it can serve any of you requests (which is the failover mechanism). So, once the ZFKC detects that the Standby Namenode needs to become the Active Namenode, it informs the Standby Namenode to start the services related to Active mode, which include updating the fsimage, rolling over the edit logs etc. This operation might take some time. Hence, the delay you mentioned.

Killing of StandBy Namenode does not affect the Active Namenode, hence you don't see any delay in performing the hdfs operations.

Hope this explains. Let me know if you have more doubts. Thanks!

View solution in original post

3 REPLIES 3

avatar
@NLAMA

There might be many reasons for the slow response.

1.Namenode metadata size.

2. Namenode Disk rpm when there is huge metadata.

3. Hardware configuration(RAM, disk mount etc.,)

4. Journal Nodes placement and ZKFC also play a key role in failover.

5. Cache Management

I suggest to work with Hortonworks Support to get the perfect route cause that imporves your performance.

avatar
New Contributor

Let me explain you my environment. There are three servers all of them acting as datanode and first two nodes are assigned as Namenode (Active and Standby). After adding some 10 MB of file in HDFS, I want to test the failover but as I explained above the response is slow around 1m 35Sec. But in case of shutting down standby namenode respond is within 2-4 seconds.

avatar
@NLAMA

If you shutdown the Active Namenode, the Standby NameNode needs to become the Active NameNode, before it can serve any of you requests (which is the failover mechanism). So, once the ZFKC detects that the Standby Namenode needs to become the Active Namenode, it informs the Standby Namenode to start the services related to Active mode, which include updating the fsimage, rolling over the edit logs etc. This operation might take some time. Hence, the delay you mentioned.

Killing of StandBy Namenode does not affect the Active Namenode, hence you don't see any delay in performing the hdfs operations.

Hope this explains. Let me know if you have more doubts. Thanks!