From: shawn@skt.mine.nu
Date: Fri Oct 17 2003 - 00:34:18 EDT
I hear of larger NFS implementations than I have, so I am confused at
why I'm seeing this.
I have an NFS/NIS&Samba server serving about 100 NFS/NIS clients and
about 300 SMB hosts. The machine has two pci fibre cards manually load
balancing three LUNs of a SAN (eg no mpxio). The machine is an E420R
with 4G RAM.
Iostat just tells me that the machine is slightly bussier than when the
machine ran "ok" but now I am constantly seeing large system times.
# mpstat
CPU minf mjf xcal intr ithr csw icsw migr smtx srw syscl usr sys wt
idl
0 69 4 2207 991 557 1853 89 659 443 7 621 1 31 9 58
1 75 4 1904 251 122 2336 93 704 454 9 647 1 30 11 57
2 76 4 1918 216 107 2368 97 724 452 8 663 1 30 11 57
3 60 5 2084 3325 1979 4513 67 627 599 7 615 1 34 10 54
which I think is another symptom of the same problem that my users are
seeing (quote ".....[over nfs] compiles [are] very, very slow...."). I
can assure you that they weren't bad a month ago. As you can see, there
is no user time so I am thinking it might be NFS releated (as
kernel=nfs). (It also looks like CPU 3 has some long running process on
it that is causing more interupts and context switches:unrelated?
symptom?). I've been slowly seeing performance go down on this machine
but it has been as we've added more and more hosts.
Have I reached the limit of the machine? Has someone walked this road
already? My /etc/system has been tweeked but ages ago (and I will
happily provide but are a bit verbose for the list).
thanks,
shawn
_______________________________________________
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 - 23:27:18 EDT