I use fileclasses in the exact same manner mentioned - one class per
project tree, and then some general classes for file-size buckets. A couple
of tricks we use to help us with this are:

* I use Puppet to configure the server, including templating the robinhood
configuration file. I have a small script that runs as an external facter
fact (https://puppet.com/docs/puppet/6.17/external_facts.html) which is
executed every puppet run (which for our environment is every 30mins) and
this outputs the project tree structure as a structured fact that I consume
to dynamically populate the list of project tree fileclasses in the
robinhood configuration file template. This means that the fileclasses are
added to the configuration relatively shortly (<=30mins) after the creation
of the project which should mean that the fileclass can be used by
robinhood when updating entries relating to the project tree. I don't have
any empircal evidence to show how accurate this approach is long term, but
it works well enough for me. We still tend to do full scans still every 3-6
months approximately.

* As Andrew mentions, queries on the fileclasses can still take a long
time, but what I've been using is the --class-info query, which outputs the
report for all the fileclasses in one go, and then just parse that output
to produce individual metrics for each fileclass. The query takes around
20-30minutes for our environment which is still a long time, but it's good
enough for my usage which is just to get coarse per-project metrics updated
every few hours.

If there are indexes or other optimisations people have around fileclasses
to improve this however I'd love to hear about it. Admittedly now we are
moving to project quotas, the need for this system is going to be lessened
as we'll have accurate per-tree space/inode usage stats from the quota
finally.

Cheers,
Matt

On Wed, 21 Oct 2020 at 07:03, Andrew Elwell <andrew.elw...@gmail.com> wrote:

> I haven't used fileclasses for a while, but we did similar (same as
> Carsten - one class per project tree) however I noticed that any
> rbh-report queries that interrogated fileclasses took a long
> (relative) time. Those of you using fileclasses, have you added any
> extra indexes?
>
> Andrew
>
>
> _______________________________________________
> robinhood-support mailing list
> robinhood-support@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/robinhood-support
>
_______________________________________________
robinhood-support mailing list
robinhood-support@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/robinhood-support

Reply via email to