Veritas File System (VxFS) buffer cache usage increases on AIX platform after the server is running for more than 248.55 days.
| Article:TECH192575 | | | Created: 2012-07-09 | | | Updated: 2012-07-20 | | | Article URL http://www.symantec.com/docs/TECH192575 |
Problem
After an AIX server is running for more than 248.55 days, the VxFS buffer cache space may increase. Because the VxFS buffer cache uses kernel pinned memory, so the kernel pinned memory usage will increase as well.
To check the VxFS buffer cache usage , you can run the vxfs command vxfsstat below,
# vxfsstat -v -e 1 </mnt_point> | grep vxi_bcache_curkbyte
vxi_bcache_curkbyte 2054784
To check system pinned memory usage, you can run the OS command svmon below,
# svmon -G
...
size inuse free pin virtual mmode
memory 5242880 5223980 18900 1883076 4921897 Ded
pg space 4194304 1035437
work pers clnt other
pin 372469 0 697 29472
in use 729833 0 118810
...
To check how many days the server has been running, you can run the OS command uptime below,
# uptime
06:26PM up 248 days, 22:22, 4 users, load average: 1.96, 1.72, 1.69
Error
Kernel pinned memory increases.
Environment
The problem only occurs where all of the following apply:
1. VxFS version 5.0, 5.1, 5.1SP1 and 6.0 running on AIX platform without the required patches.
2. System has be running for more that 248 days.
Cause
The kernel variable "lbolt" provides the time in system clock ticks since system boot. On AIX, the lbolt is declared as "long", and clock_t data type is defined as "int". VxFS uses clock_t data type in an internal calculation to decide whether buffer can be released from the VxFS buffer cache. In this calculation the 64-bit lbolt is stored into a clock_t variable and is truncated to a negative 32-bit integer once the server is running over 248.55 days. This will lead to the increase of the buffer cache usage because the cache buffers will not be released.
Solution
Currently the fix is available in the following private hotfixes.
VxFS 5.1SP1RP2 P1 HF2 on AIX
VxFS 5.1RP1 HF3 on AIX
Please contact Symantec Technical Support if the above hotfixes are required.
A workaround is to reboot the server within 248.55 days since the last reboot.
|
|
| Source | ETrack |
| Value | 2848948 |
| Description | VxFS buff cache consumption increased significantly after running over 248 days |
Related Articles
Article URL http://www.symantec.com/docs/TECH192575
Terms of use for this information are found in Legal Notices
Thank you.