Oracle Linux: BDA Nodes Experiencing Performance Slowdowns, Network Timeouts and Command Hangs (Doc ID 2582629.1)

Last updated on SEPTEMBER 21, 2021

Applies to:

Symptoms

After a BDA patch update that upgrades Oracle Linux 6.9 64bit with UEK3 to Oracle 7.5 with UEK4, all the cluster nodes experience performance slowdowns and network timeouts or network-related issues, particularly when having a high load of work running any application.

Some of the symptoms are:

• Commands like the ones below hang or take too long to respond:
  

  # mount
  # df -h
  # id user

• System logs nfs connection failure errors on /var/log/messages file like the one below:
  

  Jul 20 05:01:46 hostname kernel: nfs: server 000.000.00.00 not responding, still trying

• Application jobs or tasks fail (particularly, BDA nodes runs Cloudera's YARN jobs).

The connection between the cluster nodes uses Infiniband network interfaces, the cluster communicates with ZFS over the Infiniband network too.

Changes

 Nodes upgraded from Oracle Linux 6.9 with UEK R3 to Oracle Linux 7.5 with UEK R4 (4.1.12-124.14.1.el7uek.x86_64).

Cause

My Oracle Support provides customers with access to over a million knowledge articles and a vibrant support community of peers and Oracle experts.