eG Help

In a typical HA cluster, two or more separate machines are configured as NameNodes. At any point in time, exactly one of the NameNodes is in an Active state, and the others are in a Standby state. The Active NameNode is responsible for all client operations in the cluster, while the Standbys are simply acting as workers, maintaining enough state to provide a fast failover if necessary.

In order for the Standby node to keep its state synchronized with the Active node, both nodes communicate with a group of separate daemons called “JournalNodes” (JNs). When any namespace modification is performed by the Active node, it durably logs a record of the modification to a majority of these JNs. The Standby node is capable of reading the edits from the JNs, and is constantly watching them for changes to the edit log. As the Standby Node sees the edits, it applies them to its own namespace. In the event of a failover, the Standby will ensure that it has read all of the edits from the JournalNodes before promoting itself to the Active state. This ensures that the namespace state is fully synchronized before a failover occurs.

If the Active node takes too long to record changes in the JournalNode machines, or if the Standby node reads edits from the JournalNode machines very slowly, the Standby node will not be able to synchronize its namespace state with that of the Active node. In the event of an Active node failure therefore, the Standby node will not be able to take over from the Active node, thereby rendering the cluster unavailable to end-users.

To ensure the high availability of the cluster, it is imperative that administrators keep tabs on how quickly edits are written to and read from JournalNode machines by the Active and Standby nodes (respectively) in the cluster, detect slowness (if any), and determine where the bottleneck is. This is what the HdpNNJrnTrnTest test does!

This test measures the time taken by the Active node in a cluster to log changes to the JournalNodes. Likewise, the test also tracks how quickly/otherwise the Standby nodes read edits from the JournalNodes, so as to synchronize with the Active node. This way, the test warns administrators of slowness in Journal transactions that may potentially impact cluster availability.

Outputs of the test : One set of the results for the Hadoop cluster being monitored

Measurement	Description	Measurement Unit	Interpretation
Journal_trans	Indicates the rate at which Journal transactions were executed.	Transactions/Sec	This represents the rate at which the Active node logged changes to its namespace in the JournalNode machines. A sudden or consistent drop in the value of this measure can indicate one of the following: The namespace of the Active node has not changed much, and hence there is nothing much to write to the JournalNodes; The Active node is sluggish in recording changes to the JournalNode machines. If the slowness is owing to the latter, it is a cause for concern and hence warrants an investigation.
Avg_journal_tran_time	Indicates the average time taken by the Journal transactions for execution.	Seconds	A sudden or steady increase in the value of this measure is worrisome, as it implies that the Active node is not logging changes to the JournalNode machines as quickly as it should. Such a delay can occur if: the ActiveNode has too many changes to log, or; the changes have to be written to many JournalNode machines; typically, to ensure that the cluster tolerates a single machine failure, a minimum of 3 JournalNode daemons are required. To increase the tolerance threshold of your cluster, you may want to increase the number of JournalNode daemons to be spawned. a latent network link connects the Active node and the JournalNode machines.
Journal_sync_rate	Indicates the rate at which the Standby nodes read edits from the JournalNode machines and synchronized with the Active node.	Sync/Sec	A sudden or consistent drop in the value of this measure can indicate one of the following: The namespace of the Active node has not changed much, and hence there are not many edits in the JournalNode machines for the Standby nodes to read; The Standby nodes are sluggish in in synchronizing with the Active node. If the slowness is owing to the latter, it is a cause for concern and hence warrants an investigation.
Avg_journal_sync_time	Indicates the average time taken by Standby nodes to read edits from the JournalNode machines and synchronize with the Active node.	Seconds	A sudden or steady increase in the value of this measure is worrisome, as it implies that the Standby node is not synchronizing with the Active node as quickly as it should. Such a delay can occur if: the Standby nodes have too many changes to read from the JournalNodes and apply, or; a latent network link connects the Standby nodes and the JournalNode machines.
Journal_trans_batched	Indicates the rate at which Journal transactions were batched in sync.	Transactions/Sec