Re: File descriptors

[Chris Slominski]

In the past (no partitions) I used a table cache of 60,000 and open_files_limit of around 150,000. With an actual 30,000 tables, my cache was oversized. I have 16 GB of memory.

A select only addresses one partitioned data table; in general I am not worried about selects. My insert rates are so high, I need to keep open tables cached. Its a shame that I can't have just cached the 'current' partition for a table because that is where are the inserts are going. I don't care if an occasional select opens all partitions, but it would be nice if most of those opened partition files would subsequently be closed due to age during cache resource management. It still may not be an issue with my memory size and Linux flexibility.

Previous thread:
The problem with retaining the disk blocks was that I could not assume similar accumulation rates of data over time,and additional control system parameters are continually being added to what I archive. When I'm running out of disk space, I need to actually free up disk space so new parameters can continue to grow and to reallocate from some parameters that may no longer be accumulating any data.