[Cluster-devel] [GFS2] default drop count for lock_dlm

[Steven Whitehouse]

Hi,

I think it would be a good plan to have a zero default (i.e. off) for
the drop_count parameter, otherwise its liable to cause a performance
problem if its set too low. There is also the question of what exactly
the "right" setting actually is, as it seems it's not easy to find the
right balance. Probably the best advice is to only set it in the case
that the DLM is taking excessive amounts of memory and then use the
guidance from cat /proc/slabinfo | grep gfs2_glock to get a rough idea
of the threshold to set it at,

Steve.




diff --git a/fs/gfs2/locking/dlm/lock_dlm.h b/fs/gfs2/locking/dlm/lock_dlm.h
index 6888bd4..d074c6e 100644
--- a/fs/gfs2/locking/dlm/lock_dlm.h
+++ b/fs/gfs2/locking/dlm/lock_dlm.h
@@ -36,7 +36,7 @@
 
 #define GDLM_STRNAME_BYTES	24
 #define GDLM_LVB_SIZE		32
-#define GDLM_DROP_COUNT		200000
+#define GDLM_DROP_COUNT		0
 #define GDLM_DROP_PERIOD	60
 #define GDLM_NAME_LEN		128