NAS Queue Is Queuing / Building Up

Products

• DX Unified Infrastructure Management (Nimsoft / UIM)

Issue/Introduction

The NAS  queue is backing up or appearing "queued up" within DX Unified Infrastructure Management (UIM). This typically occurs after upgrades or during periods of high alarm volume.

Environment

• Product: DX UIM 23.4.*
• Probe: nas (Alarm Server) any version

Cause

The default configuration for alarm processing may not be sufficient for environments with high alarm volume, leading to sequential processing bottlenecks.

Resolution

Resolution

To alleviate the queue backup, adjust the bulk size and subscription block size settings in the raw configuration.

1. Adjust Hub Probe Bulk Size:
   
   

   • Open the Hub probe Raw Configuration.
   • Navigate to the postroute section.
   • Select the queue related to the nas probe.
   • Increase the bulk_size value to 100.

2. Adjust NAS Probe Subscription Block Size:
   
   

   • Open the nas probe Raw Configuration.
   • Navigate to the setup section.
   • Add or update the key subscr_block_size and set the value to 500.

3. Apply Changes:
   
   

   • Deactivate and then Activate the nas probe to initialize the new settings.

Additional Information

Technical Considerations and Constraints

• Maintenance Window Limitation: Its use is situational. If you have a high volume of alarms and no active maintenance windows, enabling subscr_block_size is appropriate. However, it must be disabled whenever a maintenance window is in use, as the system will not honor those windows while this setting is active.
  
  See: Receiving alerts even though servers are set in Maintenance mode in DX UIM.
• Performance Bottlenecks: Although increasing the subscr_block_size allows the NAS to pull a larger number of alarms into memory earlier, it does not enhance processing speed. Because the NAS operates on a single thread, it continues to handle every alarm sequentially. While bulk mode can improve efficiency by batching multiple alarms into one database transaction to reduce commit overhead, the primary bottleneck remains the single-threaded nature of sequential alarm processing.