df alert problem

[Christopher Wells]

We have a disk alert configured to alert when the disk space usage for /quovadx fluctuates 20% or more in a one minute time span.  Lately it has begun firing several times a week.  We have logged the disk space of this directory every minute but have not noticed the 20% fluctuation in the data being recorded.  I’m wondering if anyone else has experienced this kind of fluctuation.  

Alert type:  disk % full

Delta box checked.

Source: /quovadx

Source Count:  All

Comparing: >=     20

Duration:  N minutes    1

Time window:  ******

IDE version:

Host Server: hubprd1/172.31.108.163

Current Platform:

   Java version: 1.6.0

    Java vendor: IBM Corporation

        OS type: AIX

     OS version: 6.1

        OS arch: ppc

Server Build Information:

     Version: 5.8.3.0P

        Date: Mon Jan 17 2011

        Time: 09:04:46 AM

    Platform: Windows_NT

 Java Vendor: Sun Microsystems Inc.

 JDK Version: 1.6.0_16

Swing Version: 1.6.0_16

 RMI Version: 1.6.0_16

[/img]

Christopher Wells

Topic

[Daniel Lee]

Hi Christopher, hope things are going well for you at Orlando Health.  My suspension would be either your .log/.err files in your process directories or your .msg/.idx smat files.  I’m thinking that alert to measure the delta of disk usage would fire if your % disk full was reduced by 20% as well as if it increased by 20%.  Depending on how your log files are set up, if you cycle a process the .old .log and .err files are either deleted or moved to your LogHistory directory.  When you cycle your threads the .msg and .idx files are either deleted are moved to your SmatHistory directory.

If you are set up to delete your old files that that would explain the reduction.  However, I’m not an AIX expert but I’m thinking that if your History directories are on different volume groups than /quovadx then that may cause this fluctuation as well.

Do the alerts correspond to when processes or threads are being cycled?  How many days worth of SMAT files to you retain.  Do the alerts corresponding to the time when the oldest SMAT files fall out of your rotation?  

Just some thoughts.

[Christopher Wells]

Daniel,

Just noticed your reply.  Sorry for the late response.  It turned out to be a malfunction of the monitor daemon that was setting off the alert erroneously.  Bouncing the monitor daemon fixed it.

Thanks,

Christopher.

Christopher Wells

[Author]

Replies