[Linux-cluster] optimising DLM speed?

[Alan Brown]

 > It would be really interesting how long the described backup takes 
when the gfs2 filesystem is mounted exclusively on one node without locking.

The 2 million inode system backs up in about 30 minutes when mounted 
lock_nolock (0 file incremental backup using bacula)

 > For me it looks like you're facing a similar problem with gfs2 that 
has been worked around with gfs by introducing the glock_purge 
functionality that leads to a much smaller glock->dlm->hashtable and 
makes backups and the like much faster.

Quite likely. Backup performance under GFS2 is slightly worse than with 
GFS. "ls -l" in a directory is _significantly_ worse and can take up to 
4 minutes for a directory with 4000 files onboard (Remember: This is 
with the GFS2 filesystem mounted lock_dlm on one node only!)


Compounding matters, we have a network /home - mounted on the 
fileservers and NFSv3 exported to ~150 RHEL5 desktops (Lots of small 
files, LOTS of random access).

KDE, Openoffice, Thunderbird, Mozilla are all pretty lock/cachefile 
happy and hit the network /home export fairly hard, so when there's a 
performance issue the users get pretty noisy.