john allspaw <[EMAIL PROTECTED]> wrote:
> We do this at Flickr.  tmpfs for rrds, and 10 minute rsync to
> persistant disk, and building the whole thing into the
> /etc/init.d/gmetad script.
>
> No problems, really. We've tuned the 10 minute period a bit (used to
> do it more often) but other than that, no worries: ramdisk storage
> for rrds is good!

Would you share your init.d/gmetad changes?

What do you do to ensure you don't accidentally destroy data on disk
when the tmpfs files aren't there (for example, if rsync happens to
run just before gmetad starts) ?  Do you simply not use --delete, or
do you do something to check what's in the RAM partition before
copying?

Matthew Chambers <[EMAIL PROTECTED]> wrote:
> I like this idea as SOP (it means that large grids/clusters can deploy
> ganglia without worrying as much about I/O on their gmetad nodes), perhaps
> some proposal for how to do this the "right" way is in order, and either
> make the Ganglia config do it by default or at least put it in the default
> config but commented out.

I agrre, and think it should be the standard way to run Ganglia.  It's
an easy thing not to think of when you first deploy, until a while
later you realize how I/O bound your server is, or lose a disk, and
start thinking about it.  If Ganglia came standard with a well thought
out structure for keeping the RRDs in RAM and backing up regularly to
disk and restoring from disk to RAM when necessary, it would be a big
improvement - and probably save a lot of disks :)

In the meantime, let's collect people's wisdom, and pitfalls, on how
to do this right, here on the list.
  -- Cos

-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/
_______________________________________________
Ganglia-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ganglia-general

Reply via email to