03-18-2014 05:22 AM
I'm finding that some monitoring templates/policies that were in OVO8 are not in OM9. I was told by someone at IBM that OM9 was not a direct replacement of OVO8.
Anyway, Is anyone doing INODE Utilization Monitoring? Can you share what you have done? I'm having a bit of a time with this and need some help.
03-18-2014 05:38 AM
Should be easily doable by writing a script that parses 'df' output (typical option used for it would be "-i" or "-o i").
03-18-2014 09:04 AM
OMU 9.x is the next release of OMU 8.x (or OVO 8.x). OMU 9.x comes with a different set of default policies, for example OMU 9.x comes with the Infrastructure SPI while OMU 8.x came witht the meanwhile obsolete OS SPI.
But, if you download your complete configuration from OMU 8.x and upload it in OMU 9.x in the course of the upgrade or migration (e.g. to a different system or OS), you can continue to use all the OMU 8.x templates if you want to. Long term it's of course recommended that you switch to the Infra SPI as it has more features and is still being maintained.
The templates are now called policies. They are versioned and are stored as header and body (like it was already done in OMW). And Threshold Monitors in OMU 8.x are converted to Advanced Threshold Monitors and thus look a little bit different and might have slightly different functionality. But other than that all the OMU 8.x templates that were uploaded on OMU 9.x are still working the same.
03-18-2014 09:26 AM
Folks - the InfraSPI did contain inode monitoring policy in 1 of the earlier versions. This was removed due to some complexities in setting thresholds. It is planned to be added back.
Since the question arose on using the metric FS_INODE_UTIL - yes it is definitely possible to monitor the metric and send out alerts when inode usage meets / exceeds certain threshold.
For example implementation have a look at the SI-DiskCapacityMonitor policy.
03-23-2014 04:19 PM
03-23-2014 06:21 PM
>How about ext3, jfs, zfs? There's quite a few fs types - how to determine if inode monitoring is required?
Well, you can read the documentation. :-)
I had a script to check for filesystem and inode exhaustion and I had to do special for VXFS to ignore those.
03-23-2014 07:20 PM
>> Well, you can read the documentation. :-)
Point taken - I am sure documentation will fix all the problems we get :).
No that's not my point - question really is - are there some tell-tale indications -like for instatnce - df -i will always return 100% util (or does not work at all)...
03-26-2014 09:51 PM
>are there some tell-tale indications, like for instance - df -i will always return 100% util
No, it gives reasonable values. It's just when it's full, vxfs will make more inodes.