Help Center/ MapReduce Service/ User Guide (Kuala Lumpur Region)/ FusionInsight Manager Operation Guide (Applicable to 3.x)/ Alarm Reference (Applicable to MRS 3.x)/ ALM-19007 HBase GC Time Exceeds the Threshold
Updated on 2022-08-12 GMT+08:00
View PDF
Share

ALM-19007 HBase GC Time Exceeds the Threshold

Description

The system checks the old generation garbage collection (GC) time of the HBase service every 60 seconds. This alarm is generated when the detected old generation GC time exceeds the threshold (exceeds 5 seconds for three consecutive checks by default). To change the threshold, on the FusionInsight Manager portal, choose O&M > Alarm > Thresholds > Name of the desired cluster > HBase > GC > GC time for old generation. This alarm is cleared when the old generation GC time of the HBase service is shorter than or equal to the threshold.

If the multi-instance function is enabled in the cluster and multiple HBase service instances are installed, you need to determine the HBase service instance where the alarm is generated based on the value of ServiceName in Location. For example, if the HBase1 service is unavailable, ServiceName=HBase1 is displayed in Location, and the operation object in the procedure needs to be changed from HBase to HBase1.

Attribute

Alarm ID

Alarm Severity

Automatically Cleared

19007

Major

Yes

Parameters

Name

Meaning

Source

Specifies the cluster for which the alarm is generated.

ServiceName

Specifies the service name for which the alarm is generated.

RoleName

Specifies the role name for which the alarm is generated.

HostName

Specifies the object (host ID) for which the alarm is generated.

Impact on the System

If the old generation GC time exceeds the threshold, HBase data read and write are affected.

Possible Causes

The memory of HBase instances is overused, the heap memory is inappropriately allocated, or a large number of I/O operations exist in HBase. As a result, GCs occur frequently.

Procedure

Check the GC time.

  1. On the FusionInsight Manager portal, click O&M > Alarm > Alarms and select the alarm whose ID is 19007. Then check the role name in Location and confirm the IP adress of the instance.

    • If the role for which the alarm is generated is HMaster, go to 2.
    • If the role for which the alarm is generated is RegionServer, go to 3.

  2. On the FusionInsight Manager portal, choose Cluster > Name of the desired cluster > Services > HBase > Instance and click the HMaster for which the alarm is generated to go to the Dashboard page. Click the drop-down menu in the Chart area and choose Customize > GC > Garbage Collection (GC) Time of HMaster and click OK to check whether the value of GC time for old generation is greater than the threshold (exceeds 5 seconds for three consecutive checks periods by default).

    • If yes, go to 4.
    • If no, go to 6.

  3. On the FusionInsight Manager portal, choose Cluster > Name of the desired cluster > Services > HBase > Instance and click the RegionServer for which the alarm is generated to go to the Dashboard page. Click the drop-down menu in the Chart area and choose Customize > GC > Garbage Collection (GC) Time of RegionServer and click OK to check whether the value of GC time for old generation is greater than the threshold (exceeds 5 seconds for three consecutive checks periods by default).

    • If yes, go to 4.
    • If no, go to 6.

Check the current JVM configuration.

  1. On the FusionInsight Manager portal, choose Cluster > Name of the desired cluster > Services > HBase > Configurations, and click All Configurations. In Search, enter GC_OPTS to check the GC_OPTS memory parameter of role HMaster(HBase->HMaster), RegionServer(HBase->RegionServer). Adjust the values of -Xmx and -XX:CMSInitiatingOccupancyFraction of the GC_OPTS parameter by referring to the Note.

    1. Suggestions on GC parameter configurations for HMaster
      • Set -Xms and -Xmx to the same value to prevent JVM from dynamically adjusting the heap memory size and affecting performance.
      • Set -XX:NewSize to the value of -XX:MaxNewSize, which is one eighth of -Xmx.
      • For large-scale HBase clusters with a large number of regions, increase values of GC_OPTS parameters for HMaster. Specifically, set -Xmx to 4 GB if the number of regions is less than 100,000. If the number of regions is more than 100,000, set -Xmx to be greater than or equal to 6 GB. For each increased 35,000 regions, increase the value of -Xmx by 2 GB. The maximum value of -Xmx is 32 GB.
    2. Suggestions on GC parameter configurations for RegionServer
      • Set -Xms and -Xmx to the same value to prevent JVM from dynamically adjusting the heap memory size and affecting performance.
      • Set -XX:NewSize to one eighth of -Xmx.
      • Set the memory for RegionServer to be greater than that for HMaster. If sufficient memory is available, increase the heap memory.
      • Set -Xmx based on the machine memory size. Specifically, set -Xmx to 32 GB if the machine memory is greater than 200 GB, to 16 GB if the machine memory is greater than 128 GB and less than 200 GB, and to 8 GB if the machine memory is less than 128 GB. When -Xmx is set to 32 GB, a RegionServer node supports 2000 regions and 200 hotspot regions.
      • XX:CMSInitiatingOccupancyFraction to be less than and equal to 85, and it is calculated as follows: 100 x (hfile.block.cache.size + hbase.regionserver.global.memstore.size)

  2. Check whether the alarm is cleared.

    • If yes, no further action is required.
    • If no, go to 6.

Collect fault information.

  1. On the FusionInsight Manager interface of active and standby clusters, choose O&M > Log > Download.
  2. In the Service drop-down list box, select HBase in the required cluster.
  3. Click in the upper right corner, and set Start Date and End Date for log collection to 10 minutes ahead of and after the alarm generation time, respectively. Then, click Download.
  4. Contact the O&M personnel and send the collected fault logs.

Alarm Clearing

After the fault is rectified, the system automatically clears this alarm.

Related Information

None