[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [thredds] "Too Many Open Files" Error. Dataset too big?



Hi Michael:

You should set NetcdfFileCache to be at least 100 or so less than ulimit, to allow other files to be open.

All agg files need to be touched once, but not simultaneously; they are opened and then closed.

John

On 10/25/2013 12:20 PM, Michael McDonald wrote:
John et al,

Aggregations are not supposed to leave files open. If they are, thats a bug.
The number of open files should typically be less than 1000, no matter how big 
your aggregations are. It is dependent on the file cache sizes, see:

We set our NetcdfFileCache to match what we set our ulimit max # of
open files to (1mil). The hope of this is that as many files as
possible are cached in memory by the server, for quicker catalog
scans/requests. We never even approach 1mil concurrently open files.

    <NetcdfFileCache>
     <minFiles>1000</minFiles>
     <maxFiles>1000000</maxFiles>
     <scour>24 hours</scour>
   </NetcdfFileCache>


It was also my understanding that ALL of the files in a particular
dataset scan need to be touched/opened/checked before being folded
into an aggregation (this is upon initial scan for saving into the
cache/agg folder). After this initial indexing, the average # of open
files will drop off since queries are only picking/choosing a few
files from this catalog, but you still need to have all the proper
system limits bumped up in order for this initial scan/cache to ever
successfully complete in a timely fashion. correct/incorrect?

/mike


_______________________________________________
thredds mailing list
address@hidden
For list information or to unsubscribe, visit: http://www.unidata.ucar.edu/mailing_lists/