Hello Matthew,
On Sat, 2009-10-17 at 13:26 -0500, Boehm, Matthew wrote:
> I've set up a 2 NDB node cluster and am running into a swap issue. Each machine has
> 16GB of RAM. I've dedicated 13 to DataMemory and 1 to IndexMemory leaving 2 for the OS.
> There's also a 4GB swap partition. Our monitoring system keeps alerting us to "swap usage:
> 100%" and when I login to each node and run top, I see this:
>
> Mem: 16429872k total, 16371648k used, 58224k free, 94764k buffers
> Swap: 4194296k total, 4194296k used, 0k free, 1142828k cached
>
> Can anyone give suggestions on what to do in this case? How can I find out what is
> using the RAM? Because I have LockPages turned on, the OS is swapping and using almost 6GB
> of RAM. That doesn't seem right.
If you look in your ndb_x_out.log file (where x is the node number) it
should tell you in there whether or not LockPagesInMainMemory was
successful.
Please do remember that the data nodes allocate more memory than just
the DataMemory and IndexMemory. There is memory allocated for the
various buffers and structures needed. All of which is allocated upon
start-up, data nodes allocate all memory required for the life of the
process upon initialisation. That said those should not take up
anything like 6GB with the configuration you have supplied, I would be
surprised if it hits more than a couple of hundred MB with what you have
supplied. What else runs on this server?
Running the following may give you some idea of where the RAM is going:
shell> ps auxw
Kind Regards
--
Andrew Hutchings, MySQL Support Engineer, Americas
Sun Microsystems, United Kingdom
http://www.sun.com/mysql/
