Problem and solution when the database cannot be accessed due to ping-pong effect

You can solve the problem that the database cannot be accessed due to ping-pong effect as follows:
1. Issue Description
The Oracle RAC database service of one site provides two hosts to use storage resources through the multipathing mode. LUN0 in the storage device is mapped to the two hosts but one of the two hosts cannot access the LUN.
Product and version information
S5000 series
Application server using Huawei ATAE boards
The application server runs on SUSE 9 SP3.
UltraPath for Linux V100R002C01 is used.
2. Alarm Information
None
3. Handling Process
a. Run the upadm show option command on the CLI to check whether the failover function is disabled.
# upadm show option
The following information is displayed:
maxlun = 256
maxpath = 4
maxcontroller = 8
maxarray = 30
failback_interval = 60
optimal_path_check_interval = 60
failed_path_check_interval = 30
iopolicy = round_robin
lbcontroller = off
failover = on
maxtargetid = 512
b. If failover is on, run the upadm set failover=off command to disable the failover function.
# upadm set failover=off
c. Run the upadm start updateimage command to update UltraPath configuration.
# upadm start updateimage
d. Run the upadm show option command to ensure that the failover function of UltraPath is disabled.
# upadm show option
The following information is displayed:
maxlun = 256
maxpath = 4
maxcontroller = 8
maxarray = 30
failback_interval = 60
optimal_path_check_interval = 60
failed_path_check_interval = 30
iopolicy = round_robin
lbcontroller = off
failover = off
maxtargetid = 512
----End
4. Root Cause
a. Based on log analysis, two application servers use UltraPath to switch paths to access LUN0 frequently.
b. Based on log analysis, the link status between host DB1 and controller A of the storage device is Link Down.
c. Base on log analysis, the link status between DB2 and controller A is also Link Down. The storage device LUN0 switches connections to the working controllers frequently and the database log displays I/O timeout.
Conclusion
The ping-pong effect leads to repeated switchover of the LUN's working controller and unavailable access to the database.
5. Suggestions
a. Do not map a LUN to two or more application servers.
b. If you must map a LUN to two application servers in some scenarios, install a cluster software in the application servers and configure cluster reservation.
In other scenarios, refer to this case and solve them by disabling the failover function.

Scroll to top