Dear Sun Managers,
Earlier I reported a problem with a `slow backup'. After moving the contents
of one 5Gb partition from a SunOS server to one running Solaris 2.5.1, I
noted that backup times increase from 2.5 hours to over 25 hours.
After further investigations, the problem had nothing to do with the backup
devices, or that I'd recently upgrade the operating system to Solaris 2.5.1.
It was a network problem. My Solaris 2.5.1 Ultra 2 with 100Mb/s hme interface
was `autonegotiating' 100 Mb half duplex with a switched network port that had
been fixed at 100 Mb full duplex. I guess why this happens warrants further
investigation some time, but for now the immediate problem is solved.
The good news in that my 5Gb backup now takes approximately 50 minutes.
Much appreciation and thanks go to the following:
Casper Dik <casper@holland.Sun.COM>
"D. Stewart McLeod" <stewart.mcleod@boeing.com>
bismark@alta.Jpl.Nasa.Gov (Bismark Espinoza)
Rich Pieri <rich.pieri@prescienttech.com>
who all provided valuable insights, very promptly indeed.
------------------------------------------------------------------------------
In a little more detail ...
Casper Dik suggested that I test the speed when dumping to /dev/null.
This put me on the right track ...
Casper further suggested that I
- try connecting the server and backup host via a 100Mb cross-over cable
- check the 100Mb network switch
- check that switch and systems agree on full/half duplex.
D. Stewart McLeod guessed that the 10/100 Mb card wasn't configured correctly.
He sent me the URL for Sun's AnswerBook pages
http://docs.sun.com:80/ab2
along with a few appropriate excerpts of the manual.
Bismark Espinoza suggested that I needed to add an entry for the Mammoth
exabyte in /kernel/drv/st.conf.
Rich Pieri suggested using a higher blocking factor for the tape drive, and
later suggested I check that I check that both ends of the network were
communiciating at the same duplex mode.
While I will pursue the latter two suggestions later, Casper and Stewart
got me started towards the crux of the problem. After getting the local
network technicians to check the 100Mb switched hub we verified that the
autonegotiation wasn't working.
I used the ndd utility to check the settings of the hme interface.
The commands were:
ndd /dev/hme adv_100fdx_cap
ndd /dev/hme adv_100hdx_cap
ndd /dev/hme adv_10fdx_cap
ndd /dev/hme adv_10hdx_cap
ndd /dev/hme adv_autoneg_cap
I then forced the hme interface to run at 100Mb FDX using the following:
ndd -set /dev/hme adv_100fdx_cap 1
ndd -set /dev/hme adv_100hdx_cap 0
ndd -set /dev/hme adv_10fdx_cap 0
ndd -set /dev/hme adv_10hdx_cap 0
ndd -set /dev/hme adv_autoneg_cap 0
The parameters changed by ndd remain until the next system reboot.
To make them more permanent, I added the following to my /etc/system file
set hme:hme_adv_100fdx_cap = 1
set hme:hme_adv_100hdx_cap = 0
set hme:hme_adv_10fdx_cap = 0
set hme:hme_adv_10hdx_cap = 0
set hme:hme_adv_autoneg_cap = 0
Thanks again to everyone for your help.
Regards, David
-- David Beard Phone: (+61 8) 830 35709 Computer Systems Manager FAX: (+61 8) 830 33773 Departments of Statistics, Pure & Applied Mathematics University of Adelaide South Australia, 5005 E-mail: beard@maths.adelaide.edu.au
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:12:40 CDT