Summary
With RHEL 8 kernel versions below 4.18.0-553.72.1.el8_10, processes hang in blk_mq_get_tag resulting in DB instance termination.
Symptoms
- Oracle processes hang and DB instance terminates.
Log Writer Worker processes (LGnn) are reported as blocked in /var/log/messages with messages similar to the following:
kernel: INFO: task ora_lg00_racdb1:24817 blocked for more than 120 seconds. ... kernel: task:ora_lg00_racdb1 state:D stack:0 pid:24817 ppid:1 flags:0x10000084 ... kernel: blk_mq_get_tag+0x121/0x290
- Sometimes LMHB process terminates the instance before Linux identifies the hang condition and writes anything to /var/log/messages.
Affected Products
RHEL 8 or Oracle Linux 8 with RHCK kernel and any of the products below:
- FlashGrid Cluster on AWS/Azure/GCP
- FlashGrid Server on AWS/Azure/GCP
- FlashGrid Storage Farbic
Affected Versions
RHEL/RHCK kernel versions below 4.18.0-553.72.1.el8_10.
Root Cause
Linux kernel bug.
Resolution
RHEL 8: update to kernel version 4.18.0-553.72.1.el8_10 or newer.
Oracle Linux 8: if using RHCK then update to kernel version 4.18.0-553.72.1.el8_10 or newer