If users remain logged into the system for an extended period of time, a memory leak can occur. The memory leak might decrease system performance (such as increased time to load studies, or the inability to load studies). Users should manually log out after they complete their work, or when they do not need to use the application, rather than waiting for the system’s automatic log out. This practice conserves system resources, and frees up user licenses. If necessary, enforce the policy by reducing the automatic logout time. You should also regularly monitor the system resources.
To check the page scan rate:
1. Log in to the server via any account (root access is not needed).
2. Type:
vmstat
Below is an example of the return message:
procs memory page disk faults cpu
rbw swap free re mf pi po fr dr sr dd f0 s0 - - in sy cs us sy id
000 2 140632 111008 38 14 6 25 21 0 0 25 0 0 0 378 385 287 31 3 66
The sr counter indicates the scan rate and should normally be 0. If this counter is high (for example, greater than 200) for an extended period of time, the system may not have enough RAM, which might be due to a memory leak.
NOTE: There are other possible causes, such as system stress.
To return a line of values every number of seconds, use the command with a number. For example, to return values every five seconds, type vmstat 5.
For more information about the vmstat command, type man vmstat.
System stress can also cause performance issues. The system is stressed if one or more of the following symptoms are observed for more than 30 seconds:
Total CPU load exceeds 90%
Memory page scan rate is greater than 200 pages per second
Run queue or load average is greater than 4 times the number of CPUs
Network input or output is close to 90% of adapter bandwidth
When performance is degraded, look at the various system performance indicators to determine what is causing the slow performance. If your system becomes stressed regularly, contact Customer Service for assistance.
If the system does not appear to be stressed, identify the user(s) experiencing performance issues. If only one user is affected, the cause could be insufficient RAM on the client workstation, or insufficient end-to-end bandwidth. Also check for error and warning messages in the TAMAR_LOG.