From: Roy W. Erickson (erickson@pixelmagicfx.com)
Date: Wed Oct 20 2004 - 22:37:41 EDT
Sun Managers:
I rebuilt my two v480's last month. One from Sol8 to Sol9, the other was
already at Sol9 (built by Sun Engineers).
I'm seeing an NFS issue that was not present on either machine before my
rebuild that is now present on both.
These 4x900 480's are file servers running Veritas and using 2gb FC
storage.
I run 6 dual-proc Linux nodes on NFS, 25 G5's on AFP, and 60 P4's on
CIFS. They read, manipulate, then write hundreds of 10MB files through
them every minute.
The problem is, on the NFS machines, some of the proccessed files are
corrupted towards their ends. This happens when the Sun is running a
load of 15 or higher. The AFP and CIFS files are fine under the same
loads.
------ ------- ------- ------- --------
The differences on the machine that went from SOl9 to SOl9 are, to my
knowledge, the following:
1)Veritas Foundation Suite: 3.5 --> 4.0 MP1
2)Sol9 Patch set Oct2002 --> Sol9 patch set Sep2004
3) The following /etc/system parameters NOT on the current build are
set ce:ce_taskq_disable=1
set sq_max_size=2
set tune_t_fsflushr=5
set autoup=600
set maxpgio=65536
set maxphys=8388608
set vxio:vol_maxio=2048
set vxio:vol_maxioctl=65536
(The SUN engineers left these behind on both of my earlier systems)
Do any of these look like they would solve my problem?
I was hoping to not need to make 'adjustments' on the new build, so I
left these out.
Also, could there be some other NFS parameters I need to tweak that the
Sun engineers may have changed w/o my knowing?
(In the old days I needed to increase the biod's and the like.)
Thanks!
Roy
FYI:
I find the new builds run 15% faster in the point2point connections than
before. (A 480 to a G5, using drag and drop over AFP, for example,
52MB/sec ;)
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers
This archive was generated by hypermail 2.1.7 : Wed Apr 09 2008 - 23:29:35 EDT