Hi everyone,
Following on from my post yesterday regarding dCache spawning java
processes, I have carried out some more tests of our dCache and have found
the following:
The problem of excessive memory/CPU usage appears to occur during all file
transfers into our dCache (using any level of parallelism). Once a
transfer is complete, the CPU usage returns to normal but the memory used
is not released.
The problem with the sustained RAL FTS transfers is that due to the
continuous nature of the transfers, the CPU usage is always high and after
enough time all the memory runs out. Presumably this is even the case when
just performing periodic transfers: after enough of them the system will
still run out of memory. This possibly helps to explain why we have
previously had problems of this nature with our pool node, even before RAL
started their FTS transfers.
Any information anyone has on this matter would be useful. We are
currently looking into issues regarding the memory management of
our pool node. I will keep you posted on our progress.
Thanks,
Greig
--
=======================================================================
Dr Greig A Cowan http://www.ph.ed.ac.uk/~gcowan1
School of Physics, University of Edinburgh, James Clerk Maxwell Building
DCACHE PAGES: http://www.gridpp.ac.uk/deployment/admin/dcache/index.html
=======================================================================
|