memget errors in bindlog

[D. J. Bernstein]

[Sorry for all the Cc's; I don't know which of these lists are open.]

Jim Reid writes:
> Providing enough RAM for the name server is just another example of
> capacity planning IMHO.

But we've already heard that a gigabyte of RAM wasn't enough. How much
more memory do you expect in a single machine?

Once I've decided to set aside a certain amount of memory, I want the
cache to stay within that limit. I don't want it to crash.

> Maybe, but it's hard to devise an acceptable alternative.

http://cr.yp.to/dnscache.html

When the cache fills up, dnscache smoothly discards old cache entries.
The cache size is set by the system administrator.

> Another is the cost/overhead of cache
> cleaning. This could take a long time and while it is in progress the
> name server wouldn't be able to answer queries.

dnscache doesn't have this problem. Each new cache entry displaces only
a few old cache entries, just enough to make room. There aren't any
BIND-style outages.

> processes don't get smaller when they give up data space.

Your information is obsolete. Most versions of free() these days are
able to return memory to the kernel. But I'd rather keep the cache under
control in the first place.

---Dan