bind 8.2.4: limiting used memory?
Brad Knowles
brad.knowles at skynet.be
Tue Aug 7 06:36:06 UTC 2001
At 7:38 AM +0200 8/7/01, Michael Renzmann wrote:
> The problem is: we are talking about some kind of "embedded" device.
> Well, it's not one of these hardcore embedded devices with 2 MB Ram. We
> actually have 32 MB. But there is no swap space available as the device
> boots from a read-only mounted flash. If bind catches all available
> memory for its caching this will cause problems for other tasks (the
> routing protocol daemons for example, zebra is running there).
Hmm. On an embedded device without swap and with only 32MB, I
really question the usefulness of running any kind of process there
that could take up any variable amount of RAM. A small routing
process, a small firewall, an ssh daemon, maybe a NAT as part of the
firewall, perhaps a junkbusters proxy server, and you should be
pretty full. You could probably add a DHCP server to that, but
anything more I think would be pushing it.
> This is why I want to limit the memory usage, even if it would be useful
> to give bind as much memory as possible.
Since a heavily queried nameserver can easily chew up 256MB or
more, I really don't think that you can reasonably expect to get any
kind of caching nameserver on this kind of a box. You should be able
to configure port forwarding through the firewall so that any queries
that come in to port 53 on the embedded device could be redirected to
the proper nameserver(s) at your ISP, but I wouldn't try to get
anything more than that.
> When I last tried to compile bind 9, it was terribly much bigger than
> bind 8. Is there a way to reduce its executable size footprint as well?
You should be able to build it with shared libraries, and if you
then "strip" the binaries, you should be able to get the file sizes
down to something reasonable. However, this would kill any ability
to be able to debug this process. Of course, you probably don't care
since a core file from named can easily run well north of 50MB or
more....
--
Brad Knowles, <brad.knowles at skynet.be>
H4sICIFgXzsCA2RtYS1zaWcAPVHLbsMwDDvXX0H0kkvbfxiwVw8FCmzAzqqj1F4dy7CdBfn7
Kc6wmyGRFEnvvxiWQoCvqI7RSWTcfGXQNqCUAnfIU+AT8OZ/GCNjRVlH0bKpguJkxiITZqes
MxwpSucyDJzXxQEUe/ihgXqJXUXwD9ajB6NHonLmNrUSK9nacHQnH097szO74xFXqtlbT3il
wMsBz5cnfCR5cEmci0Rj9u/jqBbPeES1I4PeFBXPUIT1XDSOuutFXylzrQvGyboWstCoQZyP
dxX4dLx0eauFe1x9puhoi0Ao1omEJo+BZ6XLVNaVpWiKekxN0VK2VMpmAy+Bk7ZV4SO+p1L/
uErNRS/qH2iFU+iNOtbcmVt9N16lfF7tLv9FXNj8AiyNcOi1AQAA
More information about the bind-users
mailing list