If the lu cache is too small threads will thrash for data including
the OI tables and other important metadata.
This patch is back-ported from the following one:
Lustre-commit:
a9c51c03eaecb5234ba754b738877323a35bed51
Lustre-change: http://review.whamcloud.com/10802
Test-Parameters: alwaysuploadlogs \
mdtfilesystemtype=zfs mdsfilesystemtype=zfs ostfilesystemtype=zfs \
testgroup=review-zfs
Signed-off-by: Nathaniel Clark <nathaniel.l.clark@intel.com>
Change-Id: Ia6c85f523219706b5d41054c869e73ed5bf3c2f5
Reviewed-on: http://review.whamcloud.com/12002
Tested-by: Jenkins
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>
Tested-by: Oleg Drokin <oleg.drokin@intel.com>
#define LU_CACHE_NR_UNLIMITED -1
#define LU_CACHE_NR_DEFAULT LU_CACHE_NR_UNLIMITED
#define LU_CACHE_NR_LDISKFS_LIMIT LU_CACHE_NR_UNLIMITED
-#define LU_CACHE_NR_ZFS_LIMIT 256
+/** This is set to roughly (20 * OSS_NTHRS_MAX) to prevent thrashing */
+#define LU_CACHE_NR_ZFS_LIMIT 10240
#define LU_SITE_BITS_MIN 12
#define LU_SITE_BITS_MAX 24