In regard to: [Networker] Merits of Linux and Solaris, Stan Horwitz said...:
>Hello all;
>
>There's some talk here at Temple about migrating from proprietary
>operating systems to open operating systems. We're seriously considering
>standardizing the University on Red Hat Linux.
>
>I am wondering if any of you who run a Legato NetWorker server on a Red
>Hat Linux platform can comment on your experience with it. Would you do it
>again if you had the choice? What's your NetWorker environment like?
We are running NetWorker 6.1.3 server on Red Hat Linux, currently 7.3 but
I'll be upgrading the box to 2.1 ES in a week or so. The hardware is a
3-year old Intel motherboard with Dual 1.0 GHz PIIIs. We have about the
same number of clients (121) at Temple, but back up less data on a nightly
basis (most of our hosts do pure incremental backups).
My experiences with NetWorker on Linux have been mixed. We previously ran
Networker 4.x, and 5.2 on OSF/1 Tru64, and really never had any problems
with the hardware or OS, though in fairness our configuration was less
complex in those days.
We've had periodic problems with the RH 7.3 Linux-based server hanging.
Learning about Linux's ``SysRq'' key sequences and Red Hat's ``netdump''
crash dump mechanism have allowed us to finally get some information
from the box when it's wedged, but it's taken us a long time to get here.
Some other stream-of-thought observations (these are MHO):
- Stay away from Red Hat ES 3, at least for now. We've had numerous
problems with hangs on ES 3, and know of other sites that have too.
We're pretty sure it has to do with the Frankenkernel nature of 3
(read the list of stuff that RH backported from the 2.6.x prerelease
kernels into the 2.4.x series, and you'll likely be shocked).
- Finding an Intel box with lots of CPU power is trivial. Finding one
with lots of I/O bandwidth is much harder. It's not enough to have
N PCI or PCI-X *slots*, if they all share a single or perhaps two busses.
Intel boxes with lots of I/O bandwidth exist, but the price often
approaches or exceeds that of comparable I/O capabilities in proprietary
hardware.
- We use Linux heavily for many other services, so management has become
"used to" spending X dollars to purchase an Intel-based Linux server.
For reasons I fail to comprehend, there seems to be a tendency to balk
at spending 2X or 4X or whatever for an appropriately equipped Intel box
for something like backups, even though the price for hardware is in the
same ballpark as proprietary hardware. This relates to the previous note,
about getting a box with appropriate I/O capabilities.
- In the "even less comprehensible" department, when we're able to
purchase the hardware for cheap and get the OS for free (or cheap, in
the case of RH ES), it seems that management's inclination is to not
be willing to pay for any kind of software (or hardware) support.
Considering the up-front savings we're seeing by using Linux on Intel
you would think that purchasing a decent level of support would be
a no-brainer, but apparently not. We have Red Hat's intro/basic support
level for ES, and we don't have much positive to say about it.
I guess my main point is that switching to Linux for your backup server(s)
is great if it's for environment *standardization*, but it's a headache if
it's being done purely as an *up front* cost-saving measure. Buy good
hardware, get support on it, and don't skimp on software support either.
If management wants to skimp, tell them to skimp somewhere else.
Tim
--
Tim Mooney mooney AT dogbert.cc.ndsu.NoDak DOT edu
Information Technology Services (701) 231-1076 (Voice)
Room 242-J6, IACC Building (701) 231-8541 (Fax)
North Dakota State University, Fargo, ND 58105-5164
--
Note: To sign off this list, send a "signoff networker" command via email
to listserv AT listmail.temple DOT edu or visit the list's Web site at
http://listmail.temple.edu/archives/networker.html where you can
also view and post messages to the list.
=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=
|
