We try to retain lots, currently daily up to a few months then weekly. How we (try) to do this is by pinning the oldest Sunday (or nearest) report and allowing the system to expire the others.
It does not work though. It seems we need to manually delete each of the oldest (non pinned) reports as the system does not adhere to the 'limit' that we set in insightIQ (500 currently), our system has around 550 currently. deleting the reports via InsightIQ is painfully slow, the web page rarely returns by itself.
after a while (currently for us 550 reports) FSAnalyse jobs fail to run with the error "Cannot allocate memory" and we need to go in and remove more reports to allow the job to run again.
I'm wondering how many FSA reports other InsightIQ users keep ?
We don't have established reporting on the file data, so we don't keep just some of the numbers and delete the rest. Having the reports allows us to look back and see how and where data has changed easily (the compare function is great).
For the cost of saving the reports (few TB for a PB cluster) is an easy decision for us.
That is a very interesting retention requirement. Could you share with us why you need to keep that many FSA result set? Maybe there's a better way to get what you need.
Thank you Rob. Where are the FSA reports stored, on the cluster itself or on the IIQ appliance (which in my case uses an NFS export from Isilon for its datastore)
keep in mind, I have issues with the cluster keeping lots of reports. perhaps by the time you get that many the issues will be resolved (assuming your cluster/s get kept up to date)
For your original question -- I'll let others on the forum to chime in on how many FSA results they keep. I suspect not many customers retain as many FSA result sets as you do. Part of the reason is that on some of the larger clusters FSA jobs can take days to weeks to complete. The upcoming OneFS 8.0 release has something pretty awesome to address that. Stay tuned.
I've also filed 2 bugs against Engineering -- 1 for FSA job unable to start due to unable to allocate memory, and 1 for FSA/IIQ not expiring FSA results as specified in the "Maximum result set count".
The FSA reports that IIQ relies on always reside on-cluster, and is never part of IIQ's datastore. I don't know the reason behind the design, but I suspect security have something to do with it. InsightIQ's own datastore stores only performance-related metrics.
The FSA result set are located at -- /ifs/.ifsvar/modules/fsa/pub
The files are sqlite3 database files, one could open them up with a DB browser that reads sqlite3 files.
Example below, where the job numbers in the directory name reflects the job engine job numbers in OneFS:
[root@hop-isi-g-1 /ifs/.ifsvar/modules/fsa/pub]# ls -l
Adam_M1
1 Rookie
•
16 Posts
2404
1
Posted December 21st, 2015 15:00
Hi Rob,
OneFS 7.2.0.3
InsightIQ 3.2.2
We don't have established reporting on the file data, so we don't keep just some of the numbers and delete the rest. Having the reports allows us to look back and see how and where data has changed easily (the compare function is great).
For the cost of saving the reports (few TB for a PB cluster) is an easy decision for us.