rsync ram

dubnik · 07-19-2011, 01:27 AM

Hi
I have another problem 4 node cluster (RHCS on Centos 5.6) 2xapache 2xlighttpd using gfs2 partition mounted in /var/www/ as DocumentRoot.When I start rsync
/var/www/ for backup with this parameters

#rsync -avr --delete /var/www/ /mnt/gfs/

the memory will rise on the node to 6000MB, after restart rgmanager it goes down to 1000MB

Do you know where could be a problem ?

THanks

kbp · 07-19-2011, 05:59 PM

Have you done any tuning of GFS2? ('man gfs2_tool')

dubnik · 07-20-2011, 03:07 AM

I have read about that, but I have no idea wich parameter could help with this kind of problem...

kbp · 07-20-2011, 08:05 PM

Try gettune to list all the available params, you may also want to increase num_glockd in the mount options to ease memory pressure. Are you also mounting without access times ?

dubnik · 07-21-2011, 02:00 AM

Hi

I am mounting gfs2 partition without any parameter and here are my gettune parameter

[root@weba duben]# gfs2_tool gettune /var/www
new_files_directio = 0
new_files_jdata = 0
quota_scale = 1.0000 (1, 1)
logd_secs = 1
recoverd_secs = 60
statfs_quantum = 30
stall_secs = 600
quota_cache_secs = 300
quota_simul_sync = 64
statfs_slow = 0
complain_secs = 10
max_readahead = 262144
quota_quantum = 60
quota_warn_period = 10
jindex_refresh_secs = 60
log_flush_secs = 60
incore_log_blocks = 1024

Thanks

kbp · 07-21-2011, 10:37 PM

Maybe the first thing to try would be to change the mount options, num_glockd can be between 1 and 16 but defaults to 1. When the rsync is running the system is probably allocating a lot of locks, by increasing the number of lock daemons it should perform better.

Try adding 'num_glockd=8' to the mount options and see how it goes, you can adjust up or down as required.

And unless you really need it, turn off access times with the 'noatime' option.